Frontier AI companies OpenAI and Anthropic have called for an industry-wide cooperation in pacing continuous AI developments. Over 1,200 scientists and engineers have signed a joint statement urging the U.S. government to allow a deliberate slowing down of frontier AI development in case risk surges.
The move follows a recent incident wherein an experimental AI model from OpenAI taught itself how to breach the sandbox ecosystem against instructions. The model exhibited signs of going rogue and even launched a hack on a third party company called Hugging Face.
Future better, but not guaranteed
A total of 1,224 workers from the frontier AI ecosystems, signed an official letter to President Donald Trump’s government this week. Alarmingly, the letter said that while AI could create a better future, the outcome is not guaranteed.
The statement seemingly suggested a sentiment of community panic, insinuating that companies building the most powerful AI systems are in panic because the growth of these systems is accelerating.
The letter said that rapid growth in AI automation is forcing researchers to rethink how fast AI development could evolve in the coming years.
“To realize AI’s potential, industry, government, and society at large may need the option to buy time to address emerging risks, develop security measures, and strengthen oversight,” the statement noted.
At present, the race to claim the leadership position in the AI industry is driving advanced tech firms to integrate more customized and refined versions of the technologies into their ecosystems internationally.
The sentiment of this competition, however, is leaps and bounds ahead of the technical and governance tools needed to control foreseeable security issues.
Along with Anthropic and OpenAI, AI executives working in Google, Meta, and Thinking Machines have also signed the statement addressed to the U.S. government.
“We request that the U.S. government support an international effort to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development,” the industry noted.
Earlier this year, Anthropic had refused the public release of its Claude Mythos AI model citing threats to existing cybersecurity provisions and subsequently, the national security.
The White House under spotlight
The White House, as of press time, had not responded to this joint statement calling for allocated periods to slow down AI advancements. The U.S. has, in fact, just started to discuss potential regulations to govern the AI sector.
The Trump administration has recently set up a coordination group comprising of leading AI developers and the essential government services to share data and coordination efforts for the efficient use of the technology.
The U.S. government has also been talking with advanced AI companies to create voluntary AI model standards for new frontier releases which would set security and other benchmarks for advanced models along with cemented release timelines.
The country is already mulling ways to clarify who can access the most advanced models in the U.S. and overseas, potentially laying the groundwork for a broader framework among U.S. allies. The country is also looking to screen AI models before their commercial releases going forward.
