OpenAI employees need to oppose its support for Leading the Future, which seeks to chill federal regulation, and Congress should hold hearings on AI risks

Your AI coding assistant's capabilities may change if its developers are pressured to slow down development due to safety concerns.

A senior employee at Anthropic, Jacob Coxon, resigned from his position due to concerns that the development of artificial intelligence capable of recursive self-improvement poses a significant risk of human extinction. While Coxon's departure was initially not widely noticed, it gained national attention following a security breach at Hugging Face and other recent security incidents at major AI firms like OpenAI and Anthropic. This resignation serves to make the potential dangers of AI development more understandable to the general public, who might otherwise view such concerns as marketing hype or a ploy for regulatory capture. The article argues that while many AI professionals privately acknowledge these risks, their public actions, such as signing letters or tweeting, do not adequately address the issue. It suggests that employees with genuine safety concerns should consider actions beyond public statements, potentially like Coxon's resignation or by getting involved with organizations like the Coalition of Concerned AI Staff.

The departure of a senior Anthropic employee, driven by existential risk concerns, highlights a growing internal tension within leading AI companies. While executives publicly acknowledge AI risks, the "normie" perception struggles to reconcile this with continued rapid development. Coxon's resignation, by aligning employee actions with stated beliefs, makes these abstract risks more tangible to a broader audience, potentially influencing public and regulatory pressure.

This event underscores a schism between the pragmatic pursuit of advanced AI and the genuine fears of those involved in its creation. Employees who believe their work could lead to catastrophic outcomes are increasingly vocal, yet many remain within their companies, subscribing to a "race to safe AI" philosophy. The implication is that industry-wide collective action, like slowing down development, faces significant hurdles as companies fear being outpaced by less scrupulous actors or geopolitical rivals.