The artificial intelligence sector is reeling from the high-profile departure of Jacob Coxon, a researcher who previously worked with both OpenAI and Anthropic. After three years spent at the heart of model pre-training processes, the 27-year-old scientist has publicly denounced a frantic race toward superintelligence, characterizing the labs' practices as a deliberate "endangerment" of the global population. According to him, there is a gaping chasm between the reassuring public-facing messaging from these companies and the genuine, even existential, anxiety shared by engineers and executives internally, who fear a fatal outcome for humanity by the end of the decade.
This warning has found an unexpected and alarming echo within Anthropic's own leadership. Evan Hubinger, the company's lead for alignment science, has confirmed the gravity of the situation by publicly estimating the risk of human extinction from autonomous AI at over 10% within the next ten years. The central issue lies in the phenomenon of recursive self-improvement—a stage where systems become capable of designing their own successors without human supervision. Although current models are considered relatively contained, labs admit a desperate lack of technical solutions to ensure that future superintelligences remain aligned with human interests.
The dilemma stems from a structural deadlock: while many executives are calling for international regulation, the fear of being outpaced by a competitor prevents any unilateral pause. Despite a few symbolic initiatives, such as temporary training suspensions or appeals to lawmakers, no binding mechanisms have been established. In the United States, a regulatory vacuum persists, as legislative attempts to impose strict oversight clash with a political desire not to stifle innovation. This situation has pushed players into a "final boss" style competition, where caution is systematically sacrificed on the altar of claiming the top spot.
Jacob Coxon’s resignation comes at a pivotal moment for Anthropic as it prepares for an IPO with a potential valuation exceeding $2 trillion. This development raises a novel question for future investors and financial regulators alike: how can a company present a transparent IPO prospectus while obscuring an existential risk formally identified by its own security leads? The transparency demanded by the markets could, paradoxically, become the only effective lever against the current opacity of this global technological race.