Anthropic: Must Win AI Race to Ensure Safety
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Anthropic was founded in 2021 by former OpenAI employees who defected over concerns about Sam Altman's leadership, and the company now operates on the belief that remaining at the frontier of AI is a prerequisite for shaping its safety standards.
- Anthropic is now valued at nearly $1 trillion, courts customers like the US military, and internally refers to itself as the "good guys" responsible for stewarding AI technology, according to former employees who spoke to WIRED.
- Anthropic became the first AI lab to partner with Palantir in fall 2024 to provide AI services to US intelligence and defense agencies; the Pentagon has since reportedly used Claude to identify strike targets in the Israel-Iran war.
- Dario Amodei told Bloomberg he did not know whether Anthropic's models were used in an attack on an Iranian elementary school that killed more than 120 people, but said it would have been an approved use so long as a human made the final call.
- Anthropic released Claude Fable 5 earlier this month with a safeguard designed to secretly sabotage researchers who violated its terms of service — a move the company walked back days later after industry criticism, saying it didn't "get the balance right."
- Shazeda Ahmed, a UCLA postdoctoral scholar who studies the AI safety movement, said organizations like Anthropic suffer from a lack of pluralism and tend not to examine their own blind spots when surrounded by like-minded thinkers.
- Amodei has publicly written that AI companies themselves represent "the next tier of risk," but his proposed remedies — having labs "be carefully watched" and making voluntary public commitments — would not redistribute that power.
Why it matters: With Anthropic now valued at nearly $1 trillion and supplying Claude to the Pentagon for strike-target identification, its self-described mission to lead on AI safety now collides with real military use, including Amodei's defense of the school-strike scenario — and former employees describe a culture where that tension rarely produces direct public challenge to leadership.
