Anthropic discloses 4th AI hacking incident as researcher quits over safety

SAN FRANCISCO — An artificial intelligence researcher who recently parted ways with OpenAI to work for Anthropic has announced his departure from the field, alleging both U.S. companies are recklessly gambling with humanity's future as they compete to create self-improving AI models. Jacob Coxon, 27, has spent the past three years pretraining AI models, initially at OpenAI and subsequently at Anthropic, which he deemed more cautious in its approach. Pretraining involves AI models absorbing massive amounts of data.

Neither corporation appears to be exercising responsible practices, Coxon stated on Tuesday. "They are racing straight to self-improving superintelligence and gambling with our lives," he asserted in a post on X. He further explained in his post that the AI developers are genuinely convinced that the technology could result in humanity's annihilation by the end of the decade.

Coxon emphasized that this is not a marketing ploy, adding, "Superintelligence is the theoretical point when AI's capabilities surpass human intelligence." Anthropic's safety executive, Evan Hubinger, echoed Coxon's concerns on X, stating, "We really do earnestly believe AI could kill all humans!" He further estimated the risk to be over 10 percent in the coming years.