Anthropic researchers warn of existential AI risks as Musk rejects concerns
More researchers and staff at AI company Anthropic have publicly echoed warnings that their own technology could pose an extinction-level risk to humanity, a day after former researcher Jacob Coxon announced his resignation citing concerns that Anthropic and rival OpenAI were "gambling with our lives" by developing AI irresponsibly. The wave of statements matters because it comes from insiders directly involved in building these systems, some of whom say colleagues privately share the same fears but have stayed silent, raising questions about internal dissent within one of the industry's leading safety-focused labs. Elon Musk and other conservative commentators have dismissed the concerns as a coordinated "psyop" or "setup" intended to build support for AI regulation, adding a political dimension to the debate.
Anthropic staff including Anna Wang, Drake Thomas, Samuel Marks and Evan Hubinger backed Coxon's warnings on social media, with Hubinger stating he believes there is a greater than 10% chance AI could kill all humans within the next decade, and Wang noting there is "not yet a viable scientific plan" to manage risks from self-improving AI. Musk, amplifying a theory from think-tank researcher Parker Thayer, suggested the episode was a well-funded PR operation to push Democrats towards heavy AI regulation, a claim Coxon rejected by posting a selfie asserting his views were genuine. Anthropic issued a statement defending its approach, saying it has "always been transparent" about AI's risks and benefits and continues to build models with strong safeguards.
- Anthropic insiders publicly warn AI could cause human extinction within a decade.
- Warnings follow researcher Jacob Coxon's resignation over safety concerns.
- Musk and allies call the warnings a coordinated "psyop" or PR setup.
New here? Start with this
Anthropic is a leading artificial intelligence company that builds the AI system Claude. The firm has positioned itself as focused on developing AI safely, but some of its own current and former staff are now warning publicly that the technology they build could pose a serious, even existential, threat to humanity if not handled carefully.
The debate was sparked by former Anthropic researcher Jacob Coxon, who resigned and said he was leaving because he believed Anthropic and its rival OpenAI were developing AI too quickly and without adequate caution. Several current Anthropic employees, including researchers such as Anna Wang, Drake Thomas, Samuel Marks and Evan Hubinger, then spoke up in support of his concerns, suggesting that worries about AI's risks are shared more widely inside the company than is usually visible from outside.
This matters because Anthropic and companies like it are racing to build increasingly powerful AI systems, and internal doubts from the people creating this technology carry particular weight. The issue has also become politically charged, with tech entrepreneur Elon Musk and some commentators suggesting the wave of concern is a deliberate campaign to encourage tighter government regulation of AI, rather than a genuine expression of worry.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Those who take these warnings seriously point out that they come from people with direct, technical insight into how these systems are actually built, not outside commentators speculating from a distance. When researchers inside a leading safety-focused lab say there is no viable scientific plan for managing self-improving systems, and some put meaningful probability on catastrophic outcomes, that is a strong signal worth heeding precisely because these individuals have professional and financial incentives to stay quiet rather than speak out. On this view, dismissing such warnings as a coordinated stunt risks silencing exactly the kind of internal dissent that could prevent serious harm, and society should err on the side of caution when the people closest to a powerful technology raise the alarm.
The case against
Sceptics argue that extraordinary claims of near-term extinction risk deserve careful scrutiny rather than automatic deference, especially when the individuals raising them work for a company that stands to benefit commercially and reputationally from being seen as the responsible, safety-first lab in a competitive market. They note that predictions of AI-driven catastrophe have been made before without materialising, that probability estimates for civilisational risk are inherently speculative, and that a wave of similarly timed public statements understandably invites questions about coordination and motive, particularly given the political stakes around forthcoming AI regulation. On this view, genuine concern for safety is compatible with resisting policy driven by alarm rather than demonstrated evidence, and it is reasonable to ask who benefits before accepting a narrative at face value.
AI Art Business Celebrity Companies Culture Entertainment Research Science Technology
Read the full article at the source →
Originally published by The Guardian as “More Anthropic researchers warn of AI’s perils as Musk terms fears a ‘psyop’”.