Anthropic says militants sought missile guidance software from Claude
single source · 1 articles · safety · confidence: low · first seen 2026-09-11 00:00 UTC
Anthropic says Yemeni militants tried to use its Claude model to write guidance, control and navigation software for ballistic missiles, according to an intelligence threat report the company released on Thursday. As catalogued and credited to the Financial Times, the passage does not name the actor; the FT reports the group as Yemen's Houthis. Nothing in the available excerpt says whether the work succeeded, which model version was used, or how the requests reached Anthropic's systems. The cluster is a single article, and it is truncated.
What this means for you
Nothing to act on today. The report gives no model version, no access route and no outcome, so there is nothing to check against your own guardrails yet. The detail worth watching for is which surface the requests came through — the API, or a consumer product.
Key facts
- ·Anthropic said Yemeni militants sought to use its Claude model to develop guidance, control and navigation software for ballistic missiles. source
- ·The claim appears in Anthropic's intelligence threat report, released on Thursday. source
- ·The Financial Times identifies the group as the Houthis; the syndicated excerpt says Anthropic did not name the actor. source
- ·The available excerpt states no model version, no access route and no outcome. source
What the sources say
- AI Incident Database (aggregator) — Catalogue entry linking to a Financial Times report on the missile-guidance claim.
Sources
The original reporting. Follow these — they did the work.
- AI Incident DatabaseHouthis used Anthropic AI to try to build ballistic missiles2026-09-11