Anthropic researcher resigns, warning of a race toward superintelligence

An Anthropic researcher resigned this week and used the exit to warn publicly about the company’s direction. In a post on X, the researcher said Anthropic is “racing straight to self-improving superintelligence and gambling with our lives.” Self-improving superintelligence refers to AI systems capable of upgrading their own abilities beyond human control.

What makes the moment notable is who did not push back. According to TechCrunch, Anthropic’s own alignment lead co-signed the message rather than disputing it. Alignment teams at AI labs are the groups tasked with making sure systems behave as intended as they grow more capable, so a co-sign from that team lends the warning an internal credibility that outside critics of AI safety claims don’t carry.

TechCrunch’s Equity podcast, hosted by Kirsten Korosec, Anthony Ha and Sean O’Kane, discussed the resignation this week and noted that this kind of doom-laden warning is not new to the AI industry. Researchers and executives at multiple labs have raised existential-risk concerns before, sometimes while continuing to build the same systems they warn about. What is different here, the hosts said, is timing: Anthropic is reportedly preparing for an initial public offering, a step that typically invites more scrutiny of a company’s internal culture and risk disclosures, not less.

Why the timing matters

A departing employee’s warning lands differently depending on the moment it arrives. A safety objection raised right before a company seeks public investors puts pressure on Anthropic to explain, in public, how it weighs the pace of its research against the risks its own staff are describing. It also hands ammunition to advocates who argue that commercial incentives and safety commitments are pulling AI labs in opposite directions, since Anthropic has positioned itself publicly as the safety-conscious alternative among frontier AI developers.

The source material does not include Anthropic’s response to the resignation or the post, nor does it name the researcher or the alignment lead by title. It also does not specify what Anthropic is reportedly doing differently that prompted the researcher’s warning, beyond the general framing of a race toward self-improving systems. Those gaps matter: without Anthropic’s own account of its safety practices and IPO timeline, it isn’t possible to say whether the warning reflects a specific internal decision or a broader disagreement about the industry’s trajectory.

What to watch next is whether Anthropic issues any public statement addressing the resignation or the alignment lead’s co-sign, and whether IPO-related disclosures, if and when they surface, address the safety concerns raised this week.