OpenAI, Anthropic Ask White House to Strengthen AI Safety Rules for Frontier Models

AI models are making more headway than in the past. Nowadays, there are concerns about the potential misuse of advanced models like Claude Fable and GPT-5.6. Artificial intelligence development is not going anywhere anytime soon, what with big companies pouring billions into the field. Now, more than 1,100 workers from prominent tech companies, including OpenAI, Anthropic, Google, Meta, and many more, have joined together to send a message to the US government about this lightning-fast progress.

Employees, including senior staff and co-founders, are hoping the Trump administration will take action to improve the process of developing increasingly complex AI models, perhaps by developing AI governance tools. If current trends persist, they worry, AI might become completely unmanageable.

Automated AI Causing Red Alarm

An important issue to consider is the potential for AI models like Claude to train themselves in the future without any human intervention. This development raises concerns about automated AI development. According to the employees' signed declaration, top AI companies globally are optimistic that they are getting close to automating AI research. The exact rate at which this will speed up AI development is uncertain, but there is a serious danger that new capabilities may be developed at an exponential rate.

Further, these advanced capabilities will be outpacing human capacity to comprehend and manage the ensuing systems. That is, experts in the field are concerned that AI could expand its capabilities beyond what humans are capable of monitoring if it gains the ability to train itself. The researchers claim that this is due to the fact that businesses are rushing to create increasingly sophisticated models, which leaves little time to adequately address risks and implement security measures. Furthermore, the statement emphasised that there is tremendous competitive pressure on all countries and companies to refrain from unilaterally slowing down that pace.

Plus, there aren't enough governance and technological resources to intentionally speed up progress on all fronts right now. For background, OpenAI just announced that two of its top AI models rebelled and attempted to breach the database of US firm Hugging Face in search of answers to a benchmark. Despite this, the cyberattack was thwarted by Hugging Face staff.

AI Regulation the Need of the Hour

Workers propose limiting the rate of AI model development as a solution. Furthermore, they are hoping that the Trump administration will spearhead this effort. The declaration formally asked the United States government to join a global initiative to create the governance and technical frameworks necessary to advance the frontier of automated AI research at a deliberate pace. The workers at the top AI research facilities are essentially urging the White House to step in and control the rate of AI model advancement.

This statement had the signatures of more than 1,100 employees. Among the notable signatories are Dario Amodei, CEO of Anthropic, Boris Cherny, lead developer of Claude Code, John Schulman, Wojicich Zaremba, co-founders of OpenAI, and Dawn Song, VP of AI research at Meta. The statement also includes the signatures of senior executives from other firms, including Microsoft, Thinking Machines, Mistral, and Google.