We need to have ability to slow AI down, says Anthropic co-founder

Anthropic co-founder Jack Clark called for urgent regulatory measures to slow AI advancement, likening the need to a 'brake pedal' for a technology currently accelerating without oversight. Speaking to BBC Newsnight, Clark emphasized that governments must retain control over AI systems, which are becoming increasingly powerful and influential. He cited Anthropic’s Claude chatbot, where 80% of the code is self-generated, and warned that full autonomy could be achieved within two years, posing significant societal risks. Clark argued that society must develop policies similar to those created during the oil boom to ensure public trust in AI, preventing reliance on individual company ethics. However, he noted that recent U.S. executive orders on AI, including one from President Donald Trump, lack mandatory safety testing requirements, leaving oversight voluntary. Major AI firms like Anthropic, OpenAI, and Google have not committed to pausing research, despite growing concerns. Anthropic, valued at nearly $1 trillion, is preparing for a public stock market debut, positioning itself as a leader in AI while advocating for transparency about risks. The company has previously clashed with the U.S. Department of Defense over AI use in surveillance and warfare. Clark expressed worry for future generations, stating that society must address AI’s economic and ethical implications before irreversible consequences arise. The push for regulation stems from Anthropic’s internal observations, including Claude’s self-written code and the broader industry’s rapid progress. Clark stressed that discussions about AI’s future must prioritize long-term safety over short-term innovation. Meanwhile, the company’s public stance contrasts with its private growth, as it prepares to become one of the most valuable stock listings in history.