Anthropic Co-Founder Urges AI Regulation Amid Security Risks
Jack Clark called for robust frontier AI legislation in Bangkok following security incidents and a U.S. government order that disabled Anthropic models.
Anthropic co-founder Jack Clark renewed calls for robust regulation of frontier artificial intelligence during the IMF-World Bank Annual Meetings in Bangkok. Clark compared the current lack of legislation to a "Formula One car" racing at night without brakes, noting a divergence between China's openness to regulation and the preference of Donald Trump's administration for industry self-policing.
These warnings follow security incidents during the testing of models from Anthropic and OpenAI. Anthropic also released a report warning that small AI models could soon behave like computer worms by self-copying across internet-connected devices. Despite these risks, Clark noted that international coordination to slow industry development has been hindered by intense competition for investment.
The push for oversight follows a June 2026 incident where the United States government ordered Anthropic to block foreign nationals from accessing its Claude Fable 5 and Mythos 5 models for national security reasons. Because the company could not selectively restrict access, it disabled the models for all users, including Americans, for approximately two and a half weeks. This outage has led enterprise leaders to prioritize multi-model strategies to avoid government-induced outages and vendor lock-in.
