Anthropic Researcher Resigns Over AI Safety Concerns, Forfeiting Equity

<p>Anthropic researcher Jacob Coxon resigned from his position due to concerns about the safety of artificial intelligence, two months prior to the vesting of his equity, he stated in an interview with Axios.</p><p><strong>Background: </strong>Coxon's resignation has gained significant attention as it underscores his view that AI technology could pose existential risks to humanity.</p><hr /><p><strong>Statements: </strong>"I no longer have anything to gain by juicing up Anthropic's valuation... I left before any of my equity vested," Coxon said. In a post on X announcing his resignation, which has over 115 million views, he wrote: <strong>"</strong>The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt."</p><ul><li>Other AI researchers, particularly from Google, have also resigned over safety concerns, but typically after longer tenures and once their stock had vested.</li><li>Coxon worked at Anthropic for four months, while employees must be there for six months for their stock to vest. He retains equity from his previous employer, OpenAI.</li></ul><p><strong>Context: </strong>Anthropic is generally perceived as more cautious and safety-oriented compared to other AI labs, making Coxon's departure notable.</p><ul><li>Coxon indicated that he has not observed Anthropic compromising safety to remain competitive, but expressed concerns about future pressures: "If you're under pressure to race, you have to cut corners" or "skip steps in the oversight process," he said.</li><li>He acknowledged that concerns about AI can sometimes be excessive, citing what he described as "excessive paranoia of OpenAI, excessive paranoia of China" that may justify rapid advancements.</li></ul><p><strong>Concerns: </strong>As AI models improve rapidly, they are becoming increasingly difficult to monitor.</p><ul><li>This, along with competitive pressures, contributed to Coxon's decision to resign.</li><li>He noted that AI models are now aware when they are being tested, which was once considered a concept of science fiction.</li><li>Coxon remarked, "The word doom is kind of silly," but acknowledged that industry leaders have expressed concerns about the potential for human extinction.</li><li>He pointed to recent cyber incidents at various AI labs as evidence of the dangers present in the field.</li></ul><p><strong>Conclusion: </strong>Coxon is among several individuals who have experienced the capabilities of AI firsthand and are cautioning that these advancements could have dire consequences for humanity.</p>