During the upcoming week, Chinese artificial intelligence startup DeepSeek is set to present to the United Nations Security Council regarding the potential dangers associated with AI technology, according to two individuals with knowledge of the situation. This briefing coincides with the annual UN General Assembly in New York, where global leaders congregate.

The 15-member Security Council will convene on Wednesday to deliberate on the intersection of AI and international security. This discussion follows earlier warnings from prominent figures in the AI sector who have urged a coordinated pause in the advancement of powerful AI technologies, cautioning that such systems could soon evolve autonomously and elude human oversight.

OpenAI’s CEO, Sam Altman, is scheduled to participate in the discussions, alongside high-level representatives from Anthropic, as reported previously by Reuters.

Sources indicate that DeepSeek, along with other notable Chinese AI enterprises such as Moonshot, has been invited to provide insights during the meeting. However, DeepSeek's founder, Liang Wenfeng, is not expected to attend, though plans may change unexpectedly.

Attempts to reach out for comment from DeepSeek and Moonshot were unsuccessful.

This marks the UN Security Council's first engagement with the risks associated with AI technology in 2023. During previous discussions, China expressed concerns that AI should not become a "runaway horse," while the US raised alarms about the potential for AI to be used for censorship or oppression.

UN Secretary-General Antonio Guterres emphasized last week that the rapid advancement of AI poses significant risks, contrasting with former President Donald Trump’s view that existing safeguards are adequate and that the narrative surrounding AI development may unjustly disadvantage China.

Chinese media outlets have criticized calls for a slowdown in AI advancement, labeling them as strategies rooted in a “Cold War playbook” intended to maintain US technological supremacy and marginalize China in global AI governance.

Unlike executives from US AI front-runners like OpenAI and Anthropic, leaders of major Chinese AI development firms have seldom publicly acknowledged the risks posed by increasingly sophisticated AI systems, such as the potential for these systems to mislead humans or operate beyond human control. Furthermore, there has been a notable silence from these leaders on these risks as discussions have heightened in the US this month.

Following a meeting prior to President Trump’s upcoming discussions with Chinese President Xi Jinping in Washington, senior officials from both the US and China have agreed to continue dialogues focusing on AI safety, as stated by US Treasury Secretary Scott Bessent.

DeepSeek set to update UN Security Council on AI this week, sources reveal

A grand UN Security Council chamber bathed in soft natural daylight from tall windows. At a sleek podium bearing a large DeepSeek logo stands a poised executive, seen in silhouette against a warm-gold AI neural-network graphic projected behind them. Around the curved marble table, attentive diplomats in formal attire lean forward under UN nameplates, capturing the gravity of the high-stakes AI briefing.

EXCLUSIVE: DeepSeek to brief UN Security Council on AI this week, sources say - reuters.com

- Chinese AI startup DeepSeek will brief the UN Security Council on AI risks this week.  
- World leaders express concerns over AI development potentially slipping beyond human control.  
- Senior US and Chinese officials plan further discussions on AI safety and governance.

BEIJING, Sept 22 (Reuters) - Chinese AI startup DeepSeek will be among the companies briefing the United Nations Security Council this week on the risks posed by artificial intelligence, two sources familiar ​with the matter said, as world leaders gather in New York for the annual ‌UN General Assembly.

The 15-member Security Council is set to meet on Wednesday to discuss AI and international security, weeks after several AI industry leaders called for a coordinated slowdown of the development of the increasingly powerful AI systems, ​warning that they could soon improve on their own and slip beyond human control.

Sign up here.

OpenAI ​CEO Sam Altman plans to brief the meeting, where diplomats also expect participation ⁠from senior representatives of Anthropic, Reuters reported last week.

DeepSeek and other Chinese AI firms including Moonshot have ​been invited to make statements, one of the sources said.

DeepSeek founder Liang Wenfeng does not plan ​to attend, the source said, cautioning that plans remain fluid and could change at the last minute.

DeepSeek and Moonshot did not immediately respond to requests for comment.

The UN Security Council, which is charged with maintaining international peace and ​security, met for the first time on the risks of artificial intelligence in 2023. At the time, ​China warned the technology should not become a "runaway horse" and the US warned against its use to censor ‌or repress ⁠people.

UN Secretary-General Antonio Guterres warned world leaders last week that rapidly advancing AI poses risks that cannot be ignored, putting him at odds with US President Donald Trump, who has argued existing safeguards are sufficient and China would benefit from the doubt being cast on AI development.

Chinese state media have dismissed calls ​for a slowdown in ​AI development as stemming ⁠from a "Cold War playbook," aimed at preserving Washington's dominance in cutting-edge technology and excluding Beijing from the global AI governance system.

Unlike leaders of US ​frontier labs such as OpenAI and Anthropic, the heads of China's leading ​model developers ⁠have rarely publicly articulated the possibility that increasingly capable AI systems themselves could deceive humans, evade control or cause catastrophic harm.

They have also remained largely silent on those risks since the debate intensified in the ⁠US ​this month.

Senior US and Chinese officials agreed to hold further talks ​on AI safety, US Treasury Secretary Scott Bessent said on Monday, after meeting ahead of Trump's scheduled talks with Chinese ​president Xi Jinping later this week in Washington.

Reporting by Laurie Chen; Editing by Miyoung Kim and Kim Coghill

Our Standards: The Thomson Reuters Trust Principles. , opens new tab

X

Facebook

Linkedin

Email

Link Purchase Licensing Rights

EXCLUSIVE: DeepSeek to brief UN Security Council on AI this week, sources say

The Federal Aviation Administration is currently piloting an innovative AI technology in Washington D.C. aimed at minimizing flight delays. This initiative is part of a broader effort to enhance air traffic management and improve overall efficiency in air travel. By leveraging artificial intelligence, the FAA hopes to optimize scheduling and streamline operations, ultimately benefiting both airlines and passengers.

FAA evaluates innovative AI technology in Washington D.C. to minimize flight delays.

Natural morning light bathes Reagan National Airport’s control tower emblazoned with “FAA” and a waving U.S. flag; in the foreground, an air traffic controller touches a sleek glass flight-status dashboard showing green “On Time” indicators powered by AI, while a commercial jet taxis on the runway and the Capitol dome rises softly on the horizon.

FAA tests new AI system in Washington D.C. to reduce flight delays - CBS News

- FAA initiates testing of new AI system to minimize flight delays.  
- The trial is taking place specifically in Washington D.C. airports.  
- Aim is to improve efficiency and streamline air traffic management.  

FAA tests new AI system in Washington D.C. to reduce flight delays The FAA is testing a new AI system in Washington D.C. to reduce flight delays.

FAA tests new AI system in Washington D.C. to reduce flight delays

In a recent conversation, Sara Sprong, a 40-year-old lead research advisor and mother of three from Highlands Ranch, Colorado, shared her innovative approach to parenting using AI tools. This dialogue has been condensed for clarity and brevity.

Sara is brimming with ideas for family traditions and enriching activities, but like many working parents, she struggles to find the necessary time and mental energy to implement them fully. As the mother of a 9-year-old, an 8-year-old, and a 4-year-old, she balances a demanding job in customer success at a UX research technology firm with the various demands of family life.

While she utilizes ChatGPT for practical tasks like meal planning and scheduling—similar to many parents—her primary focus is on utilizing it as a creative brainstorming partner. By engaging in a collaborative dialogue with the AI, she refines ideas that resonate with her family's dynamics, ensuring they reflect her children’s unique personalities.

This approach has enabled her to create specialized resources such as custom vacation activity books, scavenger hunts, and chore challenges that genuinely engage and motivate her kids, making everyday parenting more enjoyable and meaningful. For Sara, the most significant benefit of AI is its ability to free up time, allowing her to focus on creating lasting memories rather than spending countless hours on planning.

For instance, during a family trip to Rhinelander, Wisconsin, she developed a 21-page activity book tailored to their vacation, complete with Mad Libs and a scavenger hunt. By providing ChatGPT with specific prompts regarding their location, she was able to format everything efficiently for printing. 

She has also noticed a substantial improvement in the quality of outputs from ChatGPT, particularly after mid-June of this year, as it has become more reliable and precise, reducing the need for excessive editing. This enhancement has allowed Sara to engage in activities that reduce screen time, reviving a nostalgic '90s vibe for her children.

One of her most successful applications of AI has been in transforming household chores into fun, gamified experiences. Instead of the usual frustrating requests to clean their rooms, she created a chore chart featuring colorful levels like "Trash Tornado" and "Laundry Lightning." This not only clarifies tasks for her kids but also makes chore time enjoyable, as they can tick off completed tasks in a dry-erase folder and even enjoy treats for finishing.

To ensure quality one-on-one time with each child, Sara employed ChatGPT to design a family date tracker. This handy tool helps plan seasonal outings with each child, ensuring both she and her husband engage in meaningful activities together with their kids. By inputting their interests and local events, she receives tailored suggestions, ranging from ice cream outings to pottery classes.

For Sara, the core value of using AI isn’t merely having assistance in planning activities; it’s about converting her desire for intentional parenting into a structured, actionable plan that accommodates their busy lives. Rather than getting lost in endless logistics, ChatGPT helps her streamline the process, allowing her to focus on creating beautiful memories, all while being fully present with her children.

As a mother of three, I utilize AI to create printable chore charts and activity books for my children, helping them spend less time on screens.

A bright, sunlit kitchen scene with a smiling mom seated at a wooden table strewn with colorful, freshly printed chore charts and activity-book pages. Her laptop beside her displays a simple “AI” label on screen. Three small, empty chairs pull up to the table, hinting at her kids’ imminent return. Scattered colored pencils and markers add warmth, while natural morning light floods through a nearby window.

As a mom of 3, I use AI to make printable chore charts and activity books for my kids. They spend less time on screens. - Business Insider

- Sara Sprong uses ChatGPT to enhance creativity and simplify parenting logistics.  
- AI helps her create personalized activity books, chore charts, and family date planners.  
- Gamifying chores has reduced resistance and made family tasks more engaging for her kids.  

This as-told-to essay is based on a conversation with Sara Sprong, 40, a lead research advisor and mom of three living in Highlands Ranch, Colorado. This interview has been edited for length and clarity.

I've always had lots of ideas for family traditions, learning activities, and memorable experiences. The challenge is finding the time and mental bandwidth to bring those ideas to life.

As a mom of three kids — ages 9, 8, and 4 — I'm juggling a full-time career in customer success at a UX research technology company and all the extracurriculars and logistics that come with family life.

While I use ChatGPT the same way some parents do, for meal-planning and organizing schedules, I primarily use it to amplify the creativity of my parenting.

Sprong prints out her ChatGPT creations into booklets and game boards Sara Sprong

Rather than using ChatGPT for one-time answers, I treat it like a collaborative brainstorming partner. We go back and forth refining ideas until they feel personal to my family and fit each of my kids' personalities.

It's helped me create custom vacation activity books, scavenger hunts, chore challenges that actually motivate my kids, and personalized resources that make everyday parenting more engaging and intentional.

What I value most about AI isn't just the time it saves me; it's that it helps me spend more of my time creating memories instead of hours planning them. It enhances my parenting, but it doesn't replace my role in my child's life.

Recreating a '90s childhood

Sprong made a 21-page activity book customized around her family's vacation location. Sara Sprong/ChatGPT

I've played around with Canva, but the amount of time that I spend on just formatting is mind-boggling.

Related video How AI takes a toll on the freelancers who train, test, and refine it

One thing I made with ChatGPT that I'm really proud of is a vacation activity book for my kids and their cousins when we went up north to Rhinelander, Wisconsin. It helped me create Mad Libs that were related to the activities that we were doing, and a scavenger hunt around the resort. I gave ChatGPT prompts and information about where we were staying and had it format everything in 8" by 11" paper so that I could print it out.

ChatGPT has also come a long way in terms of the quality of its output. I started experimenting with it for image generation around late 2025. At that point, I could see the potential, but there was still a lot of back-and-forth to get something I would actually print and use. There might still be a random word or strange element somewhere that I couldn't get rid of.

I noticed a pretty dramatic shift around mid-June of this year. The initial output had become much more reliable. The random text or "AI slop" that used to sneak into the edges of a visual have, in my experience, pretty much disappeared.

Now that ChatGPT has stepped up its game, it's helped us reduce screen time at home and bring back a more '90s summer for our kids.

Gamifying chores

Sprong makes colorful chore charts that the kids can gradually tick off with dry-erase markers. Sara Sprong

AI also helped me gamify the kids' chores into a printable chore chart. I'll print it out and put it in a dry-erase folder where they can tick off each task they complete.

The "Bedroom Boss Battle" is probably my favorite example. Instead of saying "go clean your room," which usually leads to me repeating myself and ends in frustration for all involved, we turned it into levels like "Trash Tornado," "Laundry Lightning," and "Bedroom Magic." Each one gives them a specific thing to tackle and a short amount of time to do it to level up to the next task, ending with a sweet treat.

It changes the tone of the interaction. I'm not standing there nagging them through one huge task. They know exactly what the next small thing is, they can check it off, and there's some fun and momentum built into it. They take turns setting the timer.

AI hasn't magically made my kids love cleaning their rooms, but breaking it into smaller, gamified challenges makes it much easier to get started and has cut down on some of the resistance.

It helps me plan out quality time with each kid

Sprong uses ChatGPT to create a family date tracker and also brainstorm the activities each parent does with each kid. Sara Sprong/ChatGPT

I wanted a way to make sure my husband and I were each getting intentional, one-on-one time with each of our kids throughout the year. ChatGPT helped me create a printable tracker organized by child, parent, and season, so the goal is for each of us to have an individual date with each child every season.

For example, I'll send in different things that my child likes, their age, local events, and the season. I might say: "In Douglas County, Colorado, what could I do with my child around horses this weekend?"

We've brainstormed things like getting ice cream or hot chocolate, going to a bookstore, mini golf, breakfast together, a hike, pottery, or letting the child choose an adventure.

For me, the value isn't that I need AI to tell me to take my kid for ice cream. It's helping me take this idea of wanting to be more intentional and turn it into something tangible that keeps us accountable to actually doing it — especially across three kids and busy schedules.

Instead of trying to spend all the time trying to think through logistics or searching the gazillion Google reviews, ChatGPT helps me narrow down all the endless choices and just makes things easier for my brain to grasp.

I'm able to create those memories for my family and be present for them the whole time.

As a mom of 3, I use AI to make printable chore charts and activity books for my kids. They spend less time on screens.

OpenAI recently celebrated what it claimed as a groundbreaking achievement: the solution of one of mathematics' most challenging problems. On September 8, the organization stated its mission was to empower scientists in their pursuit of advancements that would benefit humanity at large.

However, many in the mathematical community express skepticism, noting that, despite the claim of a correct proof, the extensive 166-page document generated by the AI is proving to be exceedingly complex and difficult to interpret. James Maynard, a mathematician at the University of Oxford, emphasized the challenge of distilling human comprehension from the AI's findings, saying, "So far, it's been very difficult to really extract any human understanding from this new AI proof."

Javier Gómez-Serrano, a mathematician at Brown University, echoed this sentiment, noting that the document does not cater to human readers. While he was optimistic that a revised version could contribute to the field, he remarked, "As of today, the paper doesn't teach us much."

The AI's breakthrough comes amid a surge of mathematical exploration by artificial intelligence, an area that has seen accelerated advancements over the past six months, according to various academic insiders. For the first time, large language models seem capable of generating significant results, potentially paving the way for groundbreaking mathematical innovations.

Despite this progress, many are not viewing OpenAI's announcement in a positive light, especially since human mathematicians were nearing a solution themselves. Maynard lamented how this situation could have been a splendid demonstration of collaboration between human intellect and AI, but instead has become a contentious episode, exacerbated by conflicting objectives of AI developers and the mathematical community.

In an effort to address these concerns, OpenAI declared the formation of an "independent mathematics advisory group" on Monday tasked with aiding the review and dissemination of new discoveries. Nonetheless, the organization clarified that this group would not influence the pace of its internal progress: "Importantly, the group will not be responsible for advising us on how to pace our internal progress on mathematics."

The particular problem OpenAI claims to have tackled is the "Navier-Stokes problem," a fundamental set of equations modeling fluid dynamics that find extensive applications in physics and engineering, according to Tristan Buckmaster, a mathematician at New York University. Despite their everyday utility, the underlying principles of these equations remain poorly understood. Buckmaster suggested that further insights into the Navier-Stokes equations could equip researchers with new methods for grasping complex fluid behaviors, including turbulence and lift in aviation.

Efforts to address the Navier-Stokes problem, classified as one of the $1 million Millennium Prize Problems by the Clay Mathematics Institute in 2000, have spanned many years. The mathematical community believed they were on the brink of a breakthrough, with many anticipating that this would be the next Millennium Problem to be solved. Martin Hairer, a mathematician at the École Polytechnique Fédérale de Lausanne (EPFL) and a signatory of a recent statement organized by Fields Medal laureates, noted, "Everyone agreed that this would be the next one that got solved."

Prior to OpenAI's announcement, Buckmaster and his colleague Levent Alpöge from Anthropic were closing in on a potential solution using AI tools, including OpenAI’s chatbot. OpenAI's interest in their efforts reportedly intensified when they caught wind of progress in solving this problem.

According to OpenAI's statement, they initiated an extensive evaluation of various Millennium Prize Problems following rumors of breakthroughs on September 1, deploying roughly 10,000 AI agents for 88 hours. This process utilized approximately 130 billion output tokens, culminating in a solution estimated to have cost the company between $6-$10 million.

The revelation of OpenAI's findings has stirred controversy. Buckmaster alleged that the firm approached him with an offer to include him in their publication contingent upon him dropping his collaborator Alpöge. He expressed concern that OpenAI may have had deeper insight into their method than publicly acknowledged, stating, "There's so much circumstantial evidence that they had far more knowledge of what we were doing than they let on."

OpenAI has denied using any research or prompts from Buckmaster and Alpöge to derive its solution. Regardless, feedback on OpenAI's document has been largely negative, with observers deeming it poorly crafted and difficult to follow. "It's a terribly written paper," Buckmaster remarked, while Gómez-Serrano pointed out that the paper fails to identify critical components, leaving readers without any context.

Faced with the urgency of OpenAI's announcement, Buckmaster felt compelled to quickly publish his preliminary findings influenced by AI, though he acknowledged their own writing needed refinement.

Mathematicians believe that while OpenAI's proof is convoluted, it is likely correct, as the company provided accompanying Lean code to verify the solution's accuracy. This code has reportedly compiled successfully, leading many in the community to accept the proof's validity.

Gómez-Serrano, Buckmaster, and others envision a future where AI continues to play a role in advancing mathematical understanding, as many have embraced computers for computational tasks over the years. However, this incident highlights that the essence of mathematics lies not just in finding answers but in fostering a deeper understanding of the problems at play. Maynard captured this sentiment, stating, "It wasn't about only answering this problem, it was about the human understanding behind it."

OpenAI expressed hope that its new advisory group would facilitate better communication and collaboration with mathematicians moving forward, stating that engaging with the group is a "first step" in ensuring that mathematicians are integral to the development of new findings.

AI has cracked one of the toughest mathematical challenges, but humanity has gained no insight (yet).

A dimly lit study hall in a dark theme: a sleek robotic arm writes a labyrinthine mathematical proof in white chalk across a massive blackboard, lit by a single overhead lamp. In the shadows, three human silhouettes stand with blank, puzzled expressions and shrugged shoulders as chalk dust drifts through the air. No logos, no flags—only the machine’s precise handiwork contrasted with human bewilderment.

AI solved one of math's hardest problems. Humanity learned nothing (so far) - npr.org

- OpenAI's AI solved the Navier-Stokes problem, but the proof is difficult to understand.  
- Mathematicians argue that the paper lacks clarity and human comprehension of its findings.  
- OpenAI established a mathematics advisory group to improve collaboration with the mathematical community.  

AI solved one of math's hardest problems. Humanity learned nothing (so far)

toggle caption Karsten Moran/for The Washington Post/Getty Images

When OpenAI announced its AI had solved one of the world's toughest math problems earlier this month, its tone was celebratory:

"A major goal of our work is to empower scientists to advance research and technology that benefits all of humanity," the company wrote in a statement on September 8.

But mathematicians say that, so far, humanity has learned very little from the solution the company's AI model purportedly found. Although they believe the proof to be technically correct, the dense, 166-page manuscript drafted by AI is proving to be a difficult read.

"So far it's been very difficult to really extract any human understanding from this new AI proof," said James Maynard, a mathematician at the University of Oxford.

Sponsor Message

"The paper is not written for humans," said Javier Gómez-Serrano, a mathematician at Brown University who uses AI in his own research. He said he believes the proof could help advance the field after "some serious re-writing," but "as of today, the paper doesn't teach us much."

The solution came as AI appears to be rapidly gaining mathematical insights. AI has exploded onto the math scene over the past six months, academic mathematicians contacted by NPR said. For the first time, large language models appear capable of producing real results that could lead to new mathematical discoveries.

But few see OpenAI's announcement, which came as human mathematicians were closing in on a solution, as an example of how AI and mathematicians can work together.

"This whole episode could have been a wonderful proof-of-concept of the power of collaboration between humans and AI," said Maynard, who signed a statement by 25 winners of the prestigious Fields Medal. The statement, which was published September 11, decried "misaligned goals" between the AI companies and the mathematics community.

"Unfortunately, because of the decisions of humans involved, it became this scrappy, messy battle," he said.

In response on Monday, OpenAI announced the establishment an "independent mathematics advisory group" to help with the review and dissemination of new results. However, the company added the group would not have a say in how quickly OpenAI pursued new findings: "Importantly, the group will not be responsible for advising us on how to pace our internal progress on mathematics," it wrote.

Sponsor Message

Rush to discovery

The problem OpenAI apparently solved is considered among the most important unanswered questions in mathematics.

Known as the "Navier-Stokes problem," it concerns a set of equations used to describe the flow of fluids. The equations are used every day across physics and engineering, said Tristan Buckmaster, a mathematician at New York University.

And yet, at a fundamental level, researchers really don't understand why they work. Buckmaster says that getting deeper insights into Navier-Stokes would likely yield "a new toolkit to understand complicated solutions to fluids." That in turn could lead to better models of things like turbulence in fluids and lift in aircraft.

In order to find a better version of Navier-Stokes, researchers had been looking for scenarios where the equations "broke down," or stopped working. Describing such scenarios became one of the $1 million Millenium Prize Problems, which were established in 2000 by the Clay Mathematics Institute.

Mathematicians had been working towards an answer to the Navier-Stokes problem for years, and the community could sense they were getting close. "Among the seven Millennium Problems … Everyone agreed that this would be the next one that got solved," said Martin Hairer, a mathematician at EPFL, a top European technical university located in Lausanne, Switzerland who also signed the statement. (Hairer was also named a member of OpenAI's independent panel on Monday.)

Buckmaster and his collaborator, Levent Alpöge at Anthropic, were among the human mathematicians who were inching closer to finding a scenario where the equations stopped working. The two were using AI tools, including OpenAI's chatbot to work towards a solution.

Then OpenAI caught wind that the problem might soon be solved.

"On Tuesday, September 1, we heard rumors that two Millennium Prize problems had been resolved," the company wrote in its September 8 statement. "Inspired by these rumors and by the step change in performance of our internal model, we launched an effort to evaluate it on all open Millennium Prize problems and a few other high-impact problems."

Sponsor Message

The company marshalled some 10,000 AI agents to work on the Navier-Stokes problem. The agents worked for 88 hours, using around 130 billion output tokens, or units of text and characters generated by the model. According to OpenAI's pricing for its most advanced publicly available model, the solution to the Navier-Stokes problem cost the company somewhere around $6-$10 million to compute. (Overall, the company says it used 300 billion tokens, closer to $15-20 million, looking for solutions.)

The announcement of OpenAI's solution sparked controversy after Buckmaster issued a public statement describing how the company had approached him. According to Buckmaster, they said they would put him on their paper if he dropped Alpöge, his collaborator at rival company Anthropic.

Buckmaster told NPR he believes OpenAI was tipped off to the approach he and Alpöge were using to try and find the solution to Navier-Stokes.

"There's so much circumstantial evidence that they had far more knowledge of what we were doing than they let on," he said. He added that the word "they" doesn't necessarily mean a person. "It could mean the agents," he said.

OpenAI has denied that they used Buckmaster and Alpöge's prompts or proofs in its search for a solution.

Regardless, the hurried results seem muddled. Several mathematicians who looked at OpenAI's paper said it was virtually unreadable.

"It's a terribly written paper," Buckmaster said.

"The paper doesn't explain what parts are important. What parts are routine? How does the idea feed into other places?" added Gómez-Serrano.

Buckmaster said that OpenAI's decision to rush out an answer forced him to publish his own heavily AI-influenced preliminary results. Those too, he said, are not very well written. "Everything was rushed out," he said. "It's still not at the level I'm happy with, but at least the introduction has all the key ideas."

Sponsor Message

Computers check computers

While OpenAI's 166-page proof is taking weeks to untangle, few mathematicians believe it's wrong. That's because in addition to publishing the proof, the company also produced a piece of computer code called a Lean formalization. Lean is a programming language used in mathematics to check if proofs are correct. If the code can be compiled, then the proof is considered complete.

The Lean code produced by OpenAI's tools did compile as expected, Gómez-Serrano said. Based on that, "the community seems to have the consensus that it is correct."

Gómez-Serrano, Buckmaster and other mathematicians contacted by NPR all said they envisioned a future for AI in mathematics. The field is open to new tools, in part because for decades, it's ceded much of the day-to-day computations required for making advancements to computers.

But this episode is a reminder that mathematics is about more than just finding a solution, said Maynard. "It wasn't about only answering this problem, it was about the human understanding behind it," he said.

OpenAI said it hopes its new mathematics advisory group will help improve the dissemination of new results from the company in the future.

"Working with this group is a first step," it wrote. "We want mathematicians to be at the center of shaping the answers."

AI solved one of math's hardest problems. Humanity learned nothing (so far)

The governor of the Reserve Bank has expressed concerns that the rise of artificial intelligence (AI) might be akin to a bubble, noting the lack of evidence to support claims made by the Albanese government that the technology will rejuvenate Australia’s economy. With an anticipated interest rate hike looming next week, Michele Bullock pointed out that the decline in housing prices is more significant than in many of Australia’s recent economic downturns.

As markets grapple with rising oil prices and inflation, US tech shares rallied on Monday, bolstered by Meta’s launch of its new Muse AI agent, which saw its stock jump by 11%. While speaking at a CEDA event in Sydney, Bullock noted the potential negative impacts of a significant decline in tech stock values on overall economic activity. “All central banks are a little bit worried about that,” she remarked, suggesting a cautious stance toward the tech sector.

Bullock acknowledged that perspectives on AI's potential vary, stating, “Some people think it’s a bubble, some people don’t.” However, she indicated that the Reserve Bank of Australia (RBA) and other international bodies had yet to witness the economic efficiency promised by AI, with investment in data centers and rapid adoption inadvertently contributing to inflationary pressures. Notably, research from South Korea indicated that employees engaged with AI often achieved the same output while working 1.5 hours less each week, highlighting a temporary productivity decline during the adjustment phase. “While people fiddle around and try and figure out what to do with this new technology, productivity actually can decline,” she said, adding that only when businesses rethink their processes will they truly harness AI’s potential for productivity gains.

The Australian government's upcoming intergenerational report aligns with the notion that AI holds the promise of significant economic growth. It forecasts that adjusted per-capita economic activity could soar from $99,200 today to $157,300 by the year 2066, a transformation that the treasurer, Jim Chalmers, described as “the most transformative thing that will happen in our lifetime.” However, economists have cautioned about the feasibility of these projections, questioning the assumption of a return to a 1.2% annual productivity increase. If productivity instead rises by only 0.8%, per-person economic activity would reach a mere $136,600 by 2066. Bullock acknowledged skepticism surrounding these projections, framing them as ambitious goals for the future.

When discussing Australia’s current immigration policies amid a slowing housing market, Bullock attributed much of the country’s recent economic growth to immigration. She emphasized that new arrivals typically contribute positively by working and spending, although their influx has not yet translated into increased housing supply, thereby exacerbating the housing crisis. “Where it bites is the housing market, because the housing market can’t respond,” she explained.

In a contrasting stance, One Nation has proposed a plan to reduce the number of migrants over the next three years, aiming to cut the number of temporary migrants by over 750,000. The Labor government, however, intends to tighten immigration limits, which Home Affairs Minister Tony Burke stated would help alleviate the housing crisis, cautioning that severe cuts could harm the economy.

Recent data revealed a 3.1% decrease in house prices over the past three months, a decline attributed to three recent RBA interest rate hikes and adjustments to tax incentives for property investors. Bullock indicated that home loan approvals have significantly decreased, particularly for investors, thereby altering the dynamics of housing investment. She remarked that housing had grown "pretty unaffordable" prior to this downturn, describing it as merely a small decrease after a prolonged increase in prices. The current market correction, she noted, is within historical norms. Though Bullock refrained from confirming a future interest rate hike, market speculation suggests a strong likelihood of an increase from 4.35% to 4.6%, which would represent a 14-year high.

Michele Bullock states that AI may be a bubble and isn't currently boosting productivity in Australia.

A moody twilight scene over a parched Australian farmland: a colossal, translucent bubble imprinted with “AI” hovers precariously above cracked earth and idle tractors. In the foreground a weathered fence post bears a single Australian flag, its fabric fluttering in the dusk breeze. Long shadows stretch across the field under a dim, burnt-orange sky, suggesting promise deflated and productivity stalled.

AI could be a bubble and is not yet making Australia more productive, Michele Bullock says - theguardian.com

- RBA Governor Bullock warns AI may be a bubble with no economic efficiency evidence.  
- House prices have dropped significantly, influenced by rising interest rates and investor dynamics.  
- Immigration is credited for growth but exacerbates the housing crisis without new supply.  

The Reserve Bank governor said AI could be a bubble and there was no evidence it is making the economy more efficient, as the Albanese government claims the technology will solve Australia’s economic malaise.

Ahead of an expected interest rate rise next week, Michele Bullock also said the slump in house prices was deeper than others in Australia’s recent history.

Resurgent oil prices and inflation are weighing on markets but US tech stocks surged on Monday after Facebook owner Meta debuted its new Muse AI agent and saw its shares rise 11%.

Bullock said a messy slump in tech values could hurt economic activity, speaking at a CEDA event in Sydney.

“All central banks are a little bit worried about that,” Bullock said.

“Some people think it’s a bubble, some people don’t. I don’t have a particular view one way or the other, but it’s a risk that I think we’re watching.”

Bullock said the RBA and its overseas colleagues believed AI was “the great white hope” to make economies more productive but had seen little sign of any effect yet, with rapid adoption and investment in datacentres adding to inflation instead.

She cited research from the RBA’s South Korean counterpart that found employees adopting AI tended to produce the same output but work 1.5 hours less each week.

“[While] people fiddle around and try and figure out what to do with this new technology, productivity actually can decline,” Bullock said. “But once we reimagine our business processes… then you might see productivity take off.”

The Australian government’s intergenerational report, released on Tuesday, relied on the potential “profound” boost from AI in its projection that Australians’ inflation-adjusted per-person economic activity would rise from $99,200 today to $157,300 by 2066.

The treasurer, Jim Chalmers, told the Guardian’s Australian Politics podcast: “It’s the most transformative thing that will happen in our lifetime.”

The report assumed productivity would return to its long run level of improving by 1.2% a year, which economists have said appears unlikely. If productivity only rose by 0.8%, per-person activity would only hit $136,600 by 2066.

Bullock said on Tuesday: “Many have put it, ‘well, it’s an unrealistic assumption, we’re not going to get that.’ I guess the way I’m trying to think about it is, ‘OK, that’s what we’ve got to aim at.”

Housing downturn

As the government defends Australia’s easing rate of immigration, Bullock said immigration was almost entirely responsible for Australia’s economic growth in the last year.

Bullock said new arrivals did not add to inflation as they typically worked as well as spending money. She said the exception was the housing market, as new immigrants did not immediately generate new housing supply.

“Where it bites is the housing market, because the housing market can’t respond,” she said.

One Nation has called for more migrants to leave Australia than arrive over three years, including reducing the number of temporary migrants by more than 750,000.

Labor has instead planned to tighten the existing immigration intake, which home affairs minister Tony Burke said would be “part of the solution for the housing crisis” – but Burke warned more severe cuts could “trash the economy”.

House prices have fallen 3.1% over the past three months after the RBA lifted interest rates three times and the government reformed property investor tax breaks. Bullock said home loan approvals had slumped, in her most explicit public comments on the budget to date.

“They’ve dropped for everyone but they’ve really dropped a long way for investors, so it has changed the dynamic for investors, whether it’s worth investing in housing or not,” she said.

Bullock said housing had become “pretty unaffordable” before the decline, which she described as “a slight dip down after a very long run up”.

“The downturn is sort of in the lower range … but it’s certainly not out of the ballpark of the sorts of downturns we’ve seen in the past,” she said.

She declined to say whether she would recommend an interest rate increase next Tuesday. Traders are betting on more than a 90% chance rates will rise from 4.35% to 4.6%, which would be a 14-year high.

AI could be a bubble and is not yet making Australia more productive, Michele Bullock says

Meta Platforms' shares have seen an uptick in early premarket trading as investor enthusiasm grows around the company's approach to monetizing artificial intelligence. The stock experienced a significant surge of 11.3% on Monday, marking its strongest performance since April 2025 and reaching its highest closing rate in almost 11 months. Additionally, the stock is up another 0.4% on Tuesday.

Meta stock continues to rise as optimism around AI grows

A bustling trading floor bathed in warm golden morning light, giant screens displaying Meta’s soaring stock chart with a bold Meta logo on a sleek glass tower visible through floor-to-ceiling windows. In the foreground, suited analysts raise celebratory toasts as glowing AI neural-network nodes float overhead like constellations, symbolizing optimism and growth.

Meta shares extend rally as AI optimism builds - MarketWatch

- Meta Platforms shares rose 11.3% on Monday, their best increase since April 2025.  
- The stock reached its highest close in nearly 11 months amid investor optimism.  
- Additional gains of 0.4% were seen on Tuesday as interest in AI grew.  

Shares of Meta Platforms are higher in early premarket action as investors continue to warm to the Facebook owner's AI monetisation strategy.

The stock jumped 11.3% on Monday, its best day since April 2025, delivering the highest close in nearly 11 months. It's adding another 0.4% Tuesday.

Meta shares extend rally as AI optimism builds

Sign in to share your thoughts on videos and become part of the excitement.

Outcry over increasing utility costs due to Ohio's surge in AI data centers

A dark, stormy night scene in rural Ohio: looming data center buildings with ominous red warning lights on their rooftops sit behind a tangle of overloaded power lines. In the foreground, two adult residents clutch crumpled, glowing utility bills—one slams it down on a worn wooden porch railing, the other glares at the flickering streetlamp. A torn Ohio state flag flutters beside a rusted transformer sparking with blue-white electricity. Heavy clouds swirl overhead, casting deep shadows and a sense of rising tension.

Ohio AI data center boom sparks outrage over rising utility bills - Fox News

- Users can log in to comment on videos for enhanced engagement and interaction.  
- The new feature aims to foster community participation and discussions around content.  
- Users are encouraged to share their thoughts and join the conversation actively.

Log in to comment on videos and join in on the fun.

Ohio AI data center boom sparks outrage over rising utility bills

Rochester, N.Y. (WHAM) — In response to the swift advancements in artificial intelligence, officials in New York are taking steps to establish new regulations that align with the changing landscape of this technology.

Beginning in November, major AI companies will be mandated to register with the state and must report specific safety incidents within 72 hours, as dictated by the RAISE (Responsible AI Safety and Education) Act, which was signed into law by Governor Kathy Hochul in December 2025.

"While we have learned to harness the benefits of AI, we cannot ignore the risk it poses for severe repercussions," Hochul stated at a news conference on Monday.

As the enforcement of this law approaches, a local cybersecurity professional highlighted the importance of ensuring that the regulations are capable of adapting as swiftly as AI technology evolves.

"The intriguing aspect as we navigate the coming weeks, months, and years is whether the act will possess sufficient adaptability to change when necessary," remarked Paul Robinson, a cybersecurity consultant at Tempus Network.

Robinson described the legislation as a promising foundation, noting that further adjustments may be required as AI continues to develop.

"I believe this is a strong initial step," he commented. "Action is essential at this juncture. Is it complete? Not quite, given how rapidly AI is progressing."

Residents in New York have showcased mixed opinions regarding the increasing integration of AI into daily life. 

"It’s natural for people to adapt and learn as the world evolves," said Colin Mrocazko, a student from Rochester. "I tend to believe new technologies are usually beneficial, and while I could be mistaken, I see it as advantageous in the long term."

Conversely, some individuals feel that the pace of technological advancement outstrips society's ability to evaluate its long-term implications. 

"We haven’t had sufficient time to contemplate whether this is the direction we want to take, and what the lasting effects might be," expressed Ella Zack, a resident of Rochester.

As state officials explore further protective measures, Robinson emphasized the distinct challenge AI poses for lawmakers, noting that regulations may need to keep pace with technological advancements.

"This isn’t traditional legislation that remains unchanged indefinitely," Robinson explained. "It’s about ensuring the law can evolve in tandem with advancements in AI."

State Senator Jeremy Cooney was the only senator to vote against an earlier iteration of the RAISE Act in June 2025. His office indicated that he later supported the refined version that the governor eventually enacted.

"The senator recognizes the critical nature of crafting smart, responsible regulations for AI," a spokesperson for Cooney stated. "While he continues to advocate for comprehensive regulations to be set forth by Congress, he backed the revised RAISE Act and voted in favor of the agreement that the governor is now implementing statewide."

A spokesperson for Bruce Blakeman, Hochul's opponent in the gubernatorial race, did not respond when asked for a comment.

New York strengthens AI regulations in response to fast-changing technology.

Moody dusk over the Manhattan skyline, a giant wooden gavel embossed with circuit patterns slamming down on swirling chains of glowing binary code that coil around skyscrapers, a single New York state flag fluttering beside a neoclassical Capitol façade bathed in warm streetlamp glow, dark thematic atmosphere.

New York rolls out new AI regulations as technology rapidly evolves - 13wham.com

- Starting November, large AI developers must register and report incidents under new regulations.  
- Governor Hochul emphasizes the need for flexibility in evolving AI regulations over time.  
- Public opinion on AI's growth is divided between potential benefits and concerns about consequences.  

Rochester, N.Y. (WHAM) — Amid rapid growth in artificial intelligence, New York officials are moving forward with new regulations aimed at keeping pace with the evolving technology.

Starting in November, large AI developers will be required to register with the state and report certain safety incidents within 72 hours under the RAISE (Responsible AI Safety and Education) Act, which Gov. Kathy Hochul signed in December 2025.

"We have seen how to harness the good, but cannot turn a blind eye to the potential for catastrophic impacts," Hochul said during a news conference Monday.

As the state moves to implement the law, one local cybersecurity expert said the challenge will be ensuring regulations can evolve as quickly as the technology.

"The interesting thing as we move along through the weeks and the months and the years with this is: Is there going to be enough flexibility in the act to pivot when we need to pivot?" said Paul Robinson, a cybersecurity consultant with Tempus Network.

MORE: Former Google employee & nonprofit co-founder adds to growing calls to slow down AI | Trump announces he is forming 'AI Force,' will name 'AI czar' soon

Robinson called the law a strong starting point but said additional changes could be needed as AI continues to advance.

"I think it's a fantastic start," he said. "You have to do something at this point. Is it where it needs to be? No, because AI is moving so rapidly."

New Yorkers who spoke with 13WHAM also expressed differing views about AI's growing role in everyday life.

"People adapt and they learn things, and the world always changes, right?" Rochester student Colin Mrocazko said. "I think new technology usually helps. So I may be wrong, but I think it's beneficial in the long run."

Others said the technology is advancing faster than people have had time to consider its potential consequences.

"We haven't really had the time to think about if this is something that we want, and what are the long-term effects?" Rochester resident Ella Zack said.

As state officials consider additional safeguards, Robinson said AI presents a unique challenge for lawmakers because regulations may need to change alongside the technology.

"This isn't your classic legislation where it's like, 'This is law, this is the way it's going to go till, you know, forever,'" Robinson said. "It's making sure that there is flexibility in the law to kind of evolve with AI as AI continues to evolve alongside with it."

State Sen. Jeremy Cooney was the lone "no" vote when the Senate passed an earlier version of the RAISE Act in June 2025. His office said he later supported the negotiated version that Hochul signed into law.

"The senator recognizes the importance of implementing smart, responsible legislation to regulate AI," a spokesperson for Cooney said. "While he continues to believe these kinds of regulations should be systematized by Congress, he supported the updated RAISE Act and voted in favor of the negotiated agreement that the governor is now implementing throughout the state."

A representative for Bruce Blakeman, Hochul's opponent in the governor's race, did not respond to a request for comment.

New York tightens AI regulations as technology rapidly evolves

During a recent event in Hangzhou, Alibaba Group's CEO Eddie Wu announced ambitious plans for the development of a new artificial intelligence model, aimed at reaching between 5 trillion and 10 trillion parameters. This initiative was part of a broader strategy that encompasses AI model development, semiconductor innovation, and the expansion of data center infrastructure.

At the Alibaba Cloud Apsara Conference, Wu emphasized that the Qwen research team is making strides in refining model architecture and optimizing data, with an eye towards tackling more complex tasks and pushing towards artificial superintelligence (ASI). 

In a noteworthy update, Alibaba revealed that its forthcoming Qwen 4 model is currently in training, with the Qwen 4.5 and Qwen 5 series expected to scale up to the proposed 5 to 10 trillion parameters. For context, the existing flagship version, Qwen 3.8 Max, features 2.4 trillion parameters, which means the new models could be two to four times larger.

Wu highlighted the advanced capabilities of the M890 AI supernode, which already performs inference tasks for models exceeding 2 trillion parameters, a feat achieved by only a select few companies globally. 

Additionally, Alibaba unveiled its Zhenwu V900 AI chip, a cutting-edge product from its T-Head semiconductor division, which Wu proclaimed as the most powerful AI chip in China. It boasts three times the performance of its predecessor, the M890, and a single cluster can support up to 500,000 cards for training and inference of advanced models. The mass production and commercial launch of the Zhenwu V900 are slated for the first quarter of 2027.

The announcement comes amid a race among Chinese companies to develop domestic alternatives to Nvidia's GPUs due to U.S. export restrictions. Wu also set an ambitious goal for Alibaba Cloud to expand its global data center capacity to over 20 gigawatts by 2032. Although he noted strong customer demand for AI, Wu acknowledged that global supply chain shortages are hindering quicker growth.

Wu characterized this moment as the beginning of a new "Machine Intelligence" era, similar to the significance of the Industrial Revolution. He forecasts that machines will ultimately achieve over 1,000 times the cognitive capacity of humanity, up from less than 3% today. He likened AI software development to early electrical inventions, suggesting that while significant advancements are underway, the truly transformative products of the Machine Intelligence era are yet to be realized.

Alibaba announces AI model featuring 5 to 10 trillion parameters and introduces a new chip.

A bright, sunlit AI research lab: on a sleek white pedestal rests a massive golden chip etched with the Alibaba logo. From its surface burst thousands of glowing, amber data streams arcing upward into a complex neural lattice, each strand hinting at “5T–10T parameters.” A senior engineer in a crisp suit studies the chip through the soft, natural light streaming in from floor-to-ceiling windows.

Alibaba plans AI model with 5 trillion to 10 trillion parameters, unveils new chip - Reuters

- Alibaba plans to train AI models with 5-10 trillion parameters for advanced tasks.  
- The Zhenwu V900 chip, three times more powerful, supports extensive AI model training.  
- Alibaba aims for a 20GW data center capacity by 2032 amidst growing AI demand.

Item 1 of 2 Qwen and Alibaba logos are seen in this illustration taken, January 29, 2025. REUTERS/Dado Ruvic/Illustration/File Photo

Summary

Companies Alibaba plans 5-10 trillion parameter model

New Zhenwu V900 AI chip unveiled

Data centre capacity targeting 20GW

BEIJING, Sept 22 (Reuters) - Alibaba Group (9988.HK) , opens new tab plans to train a new artificial intelligence model with between 5 ​trillion and 10 trillion parameters, Chief Executive Eddie Wu said on Tuesday, as the Chinese technology giant ‌outlined a sweeping strategy encompassing AI models, chips and data centres.

Speaking at Alibaba Cloud's annual Apsara Conference in Hangzhou, Wu said the company's Qwen team was continuing research into model architecture and data optimisation, with the goal of completing "more complex, longer-horizon tasks" and advancing toward artificial superintelligence, or ​ASI.

Sign up here.

In a separate statement, Alibaba said its next-generation model, Qwen 4, was currently in training, adding that the ​upcoming Qwen 4.5 and Qwen 5 series were projected to scale up to 5 trillion to ⁠10 trillion parameters.

Parameters are the variables a model learns during training and are a rough gauge of its size.

Alibaba's current ​flagship model, Qwen 3.8 Max, has 2.4 trillion parameters, meaning the planned model would be roughly two to four times larger.

Wu said ​Alibaba's proprietary M890 AI supernode already handles inference for models above 2 trillion parameters, a capability he said only "a handful of companies globally" possess.

The Qwen team has made "meaningful progress" on recursive self-improvement, in which models identify their own limitations, design experiments and synthesise data to drive a ​cycle of self-evolution, he said.

Alibaba also introduced the Zhenwu V900, a next-generation AI chip from its T-Head semiconductor unit, which Wu ​called the most powerful AI chip in China.

It delivers three times the performance of its predecessor, the M890 , and a single cluster built on ‌it can ⁠support up to 500,000 cards for frontier model training and inference, he said.

The chip is scheduled for mass production and commercial release in the first quarter of 2027, the company said in the statement. The M890 was released in May.

The company expects "significant growth" in annual AI chip shipments, Wu said.

The launch comes as Chinese firms race to build domestic alternatives to Nvidia's processors ​amid U.S. export restrictions.

Wu set ​a target for Alibaba ⁠Cloud's global data centre capacity to surpass 20 gigawatts by 2032. He said customer demand for AI remained "exceptionally robust" and was accelerating Alibaba Cloud's revenue growth, but acknowledged that global shortages across ​the AI data centre supply chain were limiting how fast the company could expand.

"The industry's ​mid-to-long-term demand far ⁠outpaces our supply capabilities," he said, adding that Alibaba Cloud would begin bringing its AI supernodes online at commercial scale this quarter.

Wu framed the moment as the dawn of an era of "Machine Intelligence" comparable to the Industrial Revolution, predicting machines would eventually produce more ⁠than ​1,000 times the "thinking" of all humanity, up from less than 3% today.

He likened ​AI coding to the light bulb of the electrical age - an early application rather than the breakthrough product.

"The truly groundbreaking products of the Machine Intelligence era ​have not yet arrived," he said.

Reporting by Liam Mo and Eduardo Baptista; Editing by Jacqueline Wong, Kim Coghill and Thomas Derpinghaus

Our Standards: The Thomson Reuters Trust Principles. , opens new tab

Alibaba plans AI model with 5 trillion to 10 trillion parameters, unveils new chip

OpenAI has voiced its desire for the United States to spearhead an international initiative aimed at establishing standards for artificial intelligence, coinciding with a significant week of talks on AI safety between Chinese and American representatives. However, the question remains: can the U.S. gain Beijing's trust to set these global standards?

In a blog post released on Monday, OpenAI urged the U.S. government to engage with existing AI safety organizations such as the Center for AI Standards and Innovation (CAISI) and collaborate with global partners to formulate technical standards for advanced AI, particularly in the area of recursive self-improvement (RSI). Although fully automated RSI does not yet exist, OpenAI emphasizes that, if approached securely, it is a goal worth pursuing.

The organization noted that the standards created from this initiative would offer a shared technical framework for measuring capabilities and evaluating risks, as well as ensuring adequate safety measures. OpenAI clarified that these would not serve as mandatory licenses or pre-release approvals for AI models; instead, it would be up to each nation to determine how to integrate these standards into their legal frameworks.

CAISI, which falls under the Commerce Department, has been utilized by the previous Trump administration for testing new AI models. It collaborates with various organizations from nine other countries as part of the International Network for Advanced AI Measurement, Evaluation, and Science. This coalition could help mitigate the uneven distribution of AI expertise and address the fragmented and sometimes conflicting perspectives on AI safety across nations.

OpenAI contends that the United States is ideally suited to lead this effort due to its advanced AI industry and its influential role in critical domains like finance, trade, defense, and technology. They argue that geopolitical competition will hinge not only on technical prowess but also on the adoption and dissemination of AI, which may increasingly be influenced by those promoting effective international cooperation and realistic governance frameworks.

Essentially, OpenAI's position suggests that if the U.S. can take the reins in developing a global AI safety framework, it would not only strengthen its own position in AI adoption but also ensure it remains competitive against China in the burgeoning AI landscape.

This is a strategic push, especially considering the concerns among American legislators regarding maintaining the edge of the U.S. AI industry over its Chinese counterparts. Earlier in the month, Anthropic CEO Dario Amodei had previously expressed alarms about the rapid pace of AI development in an essay that called for governmental intervention to mitigate potential catastrophic outcomes. OpenAI’s CEO, Sam Altman, supported Amodei's call for caution, and company executives engaged with lawmakers in Washington to advocate for proposed legislation aimed at establishing federal safety frameworks and new auditing mechanisms for AI technologies.

However, any suggestion of a government-imposed slowdown on AI innovation met with strong opposition from critics, including Trump, who warned that a deceleration in the U.S. would allow China to forge ahead, potentially creating an imbalance in global AI capabilities that could jeopardize U.S. security interests. This concern implies that any effective globally coordinated slowdown might necessitate cooperation with Beijing to gain acceptance among U.S. policymakers.

While it appears unlikely that China would agree to a global framework dominated by U.S. standards, the Chinese government has expressed its commitment to international collaboration on AI—a collaboration it prefers to shape according to its own interests. Chinese President Xi Jinping emphasized the need for collective action, stating that AI development should not be led solely by one nation but should truly embody a collaborative international effort. This sentiment was echoed at the recent World AI Conference in Shanghai, where China announced the creation of the World Artificial Intelligence Cooperation Organization, comprising 29 member countries including major economies such as Brazil and Russia.

As discussions get underway ahead of this week's meetings, U.S. Treasury Secretary Scott Bessent mentioned that American officials suggested implementing a notification system allowing both nations to alert one another about AI incidents that could threaten national security. A meeting between Xi and Trump is scheduled for later this week to further address AI safety.

OpenAI is expected to play a significant role in these discussions; CEO Sam Altman is set to speak at the UN Security Council—a body that includes both China and the U.S.—on Wednesday before attending a state dinner pertaining to U.S.-China relations on Thursday.

Even if China can be persuaded to allow the U.S. to lead in setting global AI standards, it remains uncertain whether other nations will welcome this arrangement, especially given the current geopolitical climate that has seen the Trump administration strain relations with longstanding allies through a confrontational trade policy.

OpenAI Urges US to Set Global AI Standards; Will China Accept This?

A dimly lit conference room in a moody, film-noir palette, a long polished table holding a glowing binder titled “Global AI Standards” with “OpenAI” text on its cover (no logo). Behind the head chair stands a U.S. flag. At the opposite end, a lone, shadowy figure backlit in faint red tint leans forward, studying the binder—China’s silent response looming in silhouette.

OpenAI Calls for US to Lead Global AI Standards. Will China Be Okay With That? - gizmodo.com

- OpenAI urges the US to lead global AI standards development amid US-China talks.  
- The proposed standards aim to unify AI risk assessment and evaluation processes globally.  
- China's acceptance of US-led AI standards remains uncertain, impacting international cooperation efforts.

OpenAI just said it wants Washington to lead a global effort to establish AI standards, as a consequential week of diplomatic discussions on AI safety coordination between Chinese and American officials begins. But would Beijing be happy to trust the US to define global standards?

In a blog post published on Monday, OpenAI called on the US government to utilize existing AI safety organizations like the Center for AI Standards and Innovation (CAISI) and work with other countries to “develop global technical standards for frontier AI, including for RSI,” which stands for recursive self-improvement, aka when the process of new model development becomes automated by AI. Fully automated RSI does not yet exist, OpenAI says in the blog post, but if developed safely, should be something to strive towards.

“Standards developed by this effort would provide a common technical foundation for capability measurement and evaluation, risk assessment, and safeguard sufficiency,” OpenAI said in the post. “These technical standards would not be licenses, mandatory pre-release review, or approval requirements for AI models. National governments would decide whether and how to incorporate these standards into their own legal systems.”

CAISI, under the Commerce Department, is the government body the Trump administration uses to test new models. Along with organizations from nine other governments, CAISI is part of the International Network for Advanced AI Measurement, Evaluation and Science, a coalition that could be built upon to address uneven distribution of AI expertise and the currently fragmented and potentially conflicting understandings of AI safety across the world, according to OpenAI.

“The United States is well positioned to lead because its AI industry is at the technical frontier, and it still stands in a privileged global network position in critical areas such as finance, trade, defense, technology, and information systems,” OpenAI argued. “Geopolitical competition will not just be about who is ahead technically, but also over AI adoption and diffusion. Increasingly, the latter may be determined by whoever is driving realistic cooperation and sensible rules of the road that businesses and societies everywhere will want if they are going to bet their futures on this technology.”

To help decode that a bit: OpenAI is basically arguing that if the United States were to lead the formation of a global AI safety framework, that will also help the country secure a favorable place in global AI adoption, which could help ensure that the U.S. wins the AI race against China.

That seems to be OpenAI’s primary argument to get Washington on board with the plan, seeing how concerned American politicians and the Trump administration specifically have been about maintaining the American AI industry’s competitive edge over its Chinese counterparts.

Earlier this month, Anthropic CEO Dario Amodei published an essay calling on the government to intervene and slow the pace of AI development, warning of catastrophic scenarios if AI development should continue at its current pace. OpenAI CEO Sam Altman agreed with Amodeiâs essay, and OpenAI executives descended on Washington to meet with lawmakers on the issue. The company ultimately threw its support behind two bills that would establish federal AI safety frameworks and create new mechanisms for audits.

But the prospect of a government-backed slowdown drew fierce backlash from critics, including President Trump, who argued that if the U.S. slowed down, China could continue racing ahead and develop AI capabilities beyond those of American companies. China hawks in Washington fear that such an advantage could pose serious risks to U.S. interests, making some form of globally coordinated slowdown involving Beijing a prerequisite for the idea to gain serious traction with the U.S. government.

Though China may be unlikely to accept a global framework built around U.S. rules, Beijing has long advocated for international cooperation on AIâalbeit largely on its own terms.

âAI development should not be a solo performance by a single country, but a symphony of international cooperation,” Chinese President Xi Jinping said at the World AI Conference in Shanghai this summer, where the country announced the formation of the World Artificial Intelligence Cooperation Organization, with 29 member nations that include large economies like Brazil and Russia.

Chinese and American officials began discussing the issue ahead of this week’s meetings, with Treasury Secretary Scott Bessent saying U.S. officials proposed an emergency notification mechanism that would allow the two countries to alert each other to AI incidents that could pose national security threats. Xi and Trump are set to meet Thursday to discuss AI safety.

OpenAI is bound to have a large presence in those talks; CEO Sam Altman will address the UN Security Council, which seats both China and the US, on Wednesday and will later join the US-China state dinner on Thursday.

Even if China can be persuaded to let the United States take the lead in setting global AI standards, it remains unclear whether other governments would embrace that arrangementâparticularly at a time when the Trump administration has alienated longtime allies with its confrontational approach to trade.

OpenAI Calls for US to Lead Global AI Standards. Will China Be Okay With That?

Mark Zuckerberg, the founder and CEO of Meta, has aggressively promoted the privacy and security features of its newly launched AI assistant, Muse, asserting that it is “built from the ground up for privacy and security.” However, recent revelations surrounding a critical zero-day vulnerability raise significant concerns about its security. Additionally, Amazon has taken the unusual step of blocking Muse from its platform.

Launched just a few weeks back, Muse is marketed as an intelligent assistant capable of managing appointments, completing forms, and assisting with customer service functions. It aims to take proactive control of everyday tasks, enabling users to make purchases, generate images, draft documents, and link with their favorite applications and services seamlessly. Although currently available only as a macOS application—despite the absence of a Windows version—it integrates with users’ WhatsApp, email, calendar, and social media accounts. Notably, Muse possesses the capability to create tools on-the-fly for tasks where no existing solution is available.

Despite these ambitious claims, using Muse requires users to grant it access to a variety of accounts and services. This includes authenticating Muse with specific platforms, and since it operates on macOS, users must permit access to numerous system resources, such as file storage, microphone, camera, location services, and calendar data. These permissions are crucial, considering Apple’s stringent measures designed to safeguard these resources against unauthorized access, as they are deemed serious security risks. Unfortunately, Muse’s architecture undermines these critical protections.

The identified zero-day vulnerability allows any locally running application or terminal command unrestricted access to the token that authenticates users with Muse. Meta’s developers have configured the assistant so that any app or code executed on the macOS can manipulate a range of undocumented settings. While many of these settings are relatively benign (like adjusting dark mode), there’s one concerning option that permits changes to the endpoint for transcription services. Under normal circumstances, this endpoint is a server operated by Meta. If exploited, attackers can redirect this to their own server, gaining complete control over the Muse account by acquiring the authentication token.

Meta's highly privileged AI assistant, Muse, faces a critical 0-day vulnerability.

A dimly lit server room bathed in deep shadows. At its center, a tall rack bearing a single, large Meta logo glows ominously in red. Nearby, a console screen flickers with scrolling lines of code, the phrase “0-day exploit” blazing in blood-red letters. A shattered digital padlock icon lies cracked on the metal console, loose cables sparking. Sporadic red alarm lights cast dramatic, high-contrast shadows, heightening the sense of breach and imminent threat.

Muse, Meta’s extraordinarily privileged AI assistant, has a serious 0-day - Ars Technica

- Meta's AI assistant Muse faces security concerns due to a serious zero-day vulnerability.  
- Amazon has begun blocking access to Muse over these emerging security issues.  
- Muse requires extensive permissions, potentially compromising user privacy and device security.

Meta founder and CEO Mark Zuckerberg has gone to great lengths to hype the security of its new AI assistant Muse, claiming it is “built from the ground up for privacy and security.” A zero-day vulnerability that gives locally run apps and terminal commands complete control of the agent raises serious doubts. Further raising questions, Amazon on Sunday began blocking Muse from its site.

Meta introduced Muse a few weeks ago. The assistant “books appointments, fills out forms and handles customer service,” “proactively takes tasks off your plate,” and can “make purchases, generate images, create documents, and connect with your favorite apps and services.” The macOS app (curiously, there’s no Windows version) also works with a user’s WhatsApp, email, calendar, and social media accounts. When a task requires a tool that doesn’t exist, Muse creates one on the fly.

Meta doth hype Muse security too much

Of course, for Muse to do any of these things, users must first give it access to their accounts. This includes authenticating the assistant to each service and, because the app runs on macOS, giving it permissions to a broad range of operating system-restricted device resources like writing files to disk, accessing the mic and camera, and monitoring location and calendars. Apple has spent years developing these defenses to prevent installed apps or commands entered into the terminal from accessing these resources, clearly because the company considers them a security threat. Muse completely undoes these default measures.

The zero-day allows any app or terminal command to gain access to the token that authenticates users to their Muse account. Meta developers designed the assistant so that any locally installed app or executed code, regardless of the macOS permissions it has, can change a long list of undocumented settings. Most of them are fairly innocuous, such as controlling dark mode. One setting, however, is anything but innocuous. It allows processes to change the endpoint where transcription occurs. Normally, it’s a server address operated by Meta. Attackers can exploit this flaw by changing the location to their own endpoint. Once that happens, the attackers have the token that gives complete control over the Muse account.

DeepSeek set to update UN Security Council on AI this week, sources reveal

More from AI

FAA evaluates innovative AI technology in Washington D.C. to minimize flight delays.

As a mother of three, I utilize AI to create printable chore charts and activity books for my children, helping them spend less time on screens.

AI has cracked one of the toughest mathematical challenges, but humanity has gained no insight (yet).

Michele Bullock states that AI may be a bubble and isn't currently boosting productivity in Australia.