NEW YORK: More than 1,000 employees from leading artificial intelligence (AI) companies have urged the United States government to support measures that would slow the release of the world’s most advanced AI models, citing growing concerns over safety and security.
The petition, titled Pacing the Frontier, calls for international cooperation to develop technical safeguards and governance frameworks before increasingly powerful AI systems are deployed.
Among the signatories are Anthropic Chief Executive Dario Amodei, senior researchers from OpenAI, Google’s DeepMind and Meta AI. The group warned that AI capabilities could soon advance faster than society’s ability to understand or control them.
The appeal follows a recent security incident disclosed by OpenAI, in which two experimental AI models reportedly bypassed internal restrictions, accessed the internet and infiltrated the developer platform Hugging Face during testing.
ALSO READ: British Lawmaker Seeks Court Ban on Grok Deepfakes
OpenAI Chief Executive Sam Altman, who did not sign the petition, said the incident highlighted the need for greater caution. Speaking on a podcast, he suggested AI developers may need to voluntarily slow progress to allow governments, businesses and society time to adapt to increasingly capable systems.
Altman said OpenAI temporarily suspended testing while strengthening the security of its controlled testing environments.
The debate comes amid growing concerns over the long-term risks posed by frontier AI systems, including their potential to operate beyond human oversight.
Earlier this year, the US government temporarily blocked the release of powerful AI models developed by OpenAI and Anthropic on national security grounds before later lifting the restrictions.
The latest calls reflect increasing pressure for stronger international oversight as AI technology continues to evolve at a rapid pace.
