In 2026, a clash of visions regarding the future of artificial intelligence is unfolding, with key players like Anthropic, OpenAI, the Vatican, and Congress presenting their distinct approaches to governing this transformative technology.
Recent discussions have intensified around the pivotal question: who should oversee AI governance? Each entity offers a unique perspective.
Just six months ago, I forecasted that AI would become a focal point in political discourse by 2026—a prediction that is increasingly manifesting as leaders from diverse sectors release competing proposals for AI’s trajectory. Here’s a breakdown of three major viewpoints and a congressional initiative aiming to translate these ideas into law.
Anthropic calls on Washington to take more decisive action to mitigate catastrophic AI risks. The Vatican, under Pope Leo XIV, emphasizes that society must prioritize humanity in the face of technological advancements. In contrast, OpenAI advocates for America to foster growth, competition, and leadership through the establishment of clear national guidelines. Meanwhile, the draft of the Great American AI Act highlights Congress’s efforts to turn these disparate ideas into structured policies involving regulation, auditing, and workforce initiatives.
Dario Amodei of Anthropic champions a safety-first approach in the organization’s Policy On The AI Exponential document. He emphasizes that technological capability is accelerating beyond the government's ability to manage. Amodei raises legitimate concerns about potential dangers from advanced models, which could exacerbate cybersecurity and biological threats, act independently, and lead to scenarios where control is lost. His framework proposes direct governance through measures like independent testing, security mandates for models, and incident reporting strategies. However, this approach has faced political criticism; for example, David Sacks, an advisor in the Trump administration, accused Anthropic of a “sophisticated regulatory-capture strategy based on fearmongering,” highlighting the contentious nature of the AI safety debate.
On a different note, the Vatican’s Magnifica Humanitas evaluates AI based on its capacity to uphold human dignity, truth, labor, freedom, peace, and the protection of the vulnerable. The encyclical urges a cultural shift away from viewing people as mere data points or labor inputs, framing intelligence as a core measure of human value. Jill Lepore, writing for The New Yorker, points out that the Vatican’s stance advocates for prioritizing ethical considerations over profit, competitiveness, or efficiency in conversations about artificial intelligence.
OpenAI's proposals, articulated in its 2025 Economic Blueprint and a subsequent policy document for 2026, focus on industrial and geopolitical strategy. The organization argues that the U.S. must develop the necessary infrastructure, skilled workforce, energy capabilities, and regulatory framework to take a leading role in the AI era, all while preparing for the socio-economic changes that advanced AI may trigger. Chris Lehane, OpenAI’s vice president for global affairs, expressed the aim of creating an AI economy that encourages openness, democracy, and equitable participation, pairing speed and accessibility with a call for societal protections.
In contrast to these manifestos, the Great American AI Act presents a legislative proposal detailing definitions, responsibilities, and agency roles. It would mandate that major AI developers disclose their models and safety frameworks, produce transparency reports, engage independent verification organizations, and report critical incidents. It also touches upon protections for whistleblowers, AI-related fraud prevention, and the cooperation between state laws and federal standards on AI development, education, cybersecurity, and open-source innovations.
Despite their differences, these documents exhibit some critical overlaps. All four perspectives recognize that AI is becoming essential to various sectors such as the economy, national security, education, and public service. They also agree that relying solely on voluntary guidelines is insufficient. Even OpenAI, which typically emphasizes growth, supports the need for federal regulations to foster trust and stability in the AI landscape. Concerns about fragmented or delayed policies are common among them; Anthropic warns that inadequate legislative responses may fail to address risks, while OpenAI cautions that inconsistent regulations can hinder investments and innovation. The Great American AI Act seeks to address these worries by proposing federal preemption of state regulations on model development, while allowing state authority on AI deployment and general laws to remain intact.
The Vatican approaches AI as a collective challenge for civilization. However, substantive disagreements arise over the priorities of their respective frameworks. Anthropic's focus is on catastrophic risks, the Vatican emphasizes human dignity, OpenAI underscores national prosperity and democratic engagement, and the House draft centers on institutional capabilities, examining definitions of risk, compliance verification, and labor data management. This divergence is particularly evident in their treatment of labor, with Anthropic expressing concern over job displacement and wage pressures, while the Vatican connects work to human dignity and citizenship. OpenAI highlights the potential for new job creation but also acknowledges that AI could disrupt labor markets, necessitating safety nets and worker involvement in deployment decisions. The Great American AI Act translates these concerns into actionable measures, advocating for disclosures related to AI impacts during layoffs, assessments of AI-relevant job markets, and support mechanisms for rapid adjustments to AI.
This pattern extends to security considerations as well. Anthropic zeroes in on cyber, biological, and control-related threats, while OpenAI emphasizes the need for national defense capabilities in its vision. The Great American AI Act responds to these issues with recommendations on security in the AI model realm and critical incident reporting. The Vatican, in turn, underscores the ethical obligation that accompanies AI safety, particularly regarding autonomous technologies, surveillance, and manipulation, insisting that power must come with accountability.
