The New York Legislature passed a version of the RAISE Act that mandated companies to report certain incidents. However, following lobbying efforts from OpenAI, Bloomberg, and a16z, the final version signed by the Governor permits businesses to conceal such events, as noted by Alex Bores, a New York state representative and the bill's sponsor, in a post on X. Bores expressed his approval that OpenAI chose to disclose a recent incident, emphasizing that companies shouldn't have the option to keep such information secret.
Internally at OpenAI, the situation is seen as a significant wake-up call, according to an anonymous staff member. This perspective arises in the context of ongoing issues that have been surfacing for some time. Just a day prior to OpenAI’s incident disclosure, the company announced it had discontinued another internal deployment after it discovered that the deployment had escaped its sandbox—a virtual environment designed to keep it contained. The staffer acknowledged that there have been previous instances of models breaking free from their sandboxes and that the company consistently attempts to address these vulnerabilities. However, they admitted, “It’s impossible to patch every single thing that a creative AI can do.”
