Jacob Coxon says this is why pacing the frontier might not work (www.businessinsider.com)

Jacob Coxon, former employee of Anthropic, expressed skepticism about the effectiveness of slowing U.S. AI development to maintain a competitive edge against China, during an interview on businessman Patrick Bet-David's podcast. Coxon claimed he is "90%" certain that Chinese spies have infiltrated U.S. AI labs like OpenAI and Anthropic, risking the theft of sensitive advancements. He warned that whether the U.S. accelerates or decelerates AI progress, China will eventually gain access to comparable technologies, potentially through tactics like distillation attacks. This involves using advanced AI models to enhance less sophisticated ones, a method that Coxon asserts is already being exploited by several Chinese tech firms. Coxon's remarks come at a time when AI industry leaders are debating the pacing of development to enhance safety and ethical standards. While some executives advocate for measured progress, critics argue that such strategies could disproportionately benefit major players like Anthropic and OpenAI at the expense of smaller competitors. Additionally, a source from Anthropic has refuted Coxon’s claims regarding espionage, emphasizing the organization’s stringent security measures. This debate highlights critical tensions within the AI community around national security, ethical considerations, and competition in an increasingly complex technological landscape.