Provenance record

Houthis used Anthropic AI to try to build ballistic missiles

Financial Times: Artificial Intelligence (tier 1, news) 2026-09-11T09:02:54.000Z Original ↗

Source note: Headlines and standfirsts only; full text is paywalled.

Discourse valence

-71

Strongly adverse

confidence 75% · 5 items · range -100 to -45

-100

-50

+50

+100

-71

Adverse readingFavourable reading

Consensus of 5 models from different labs. Spread 55 points, agreement low.

Cyber capability Existential and catastrophic risk Loss of control and alignment

Excerpt as ingested

Missile test by group apparently failed, but exposes limits of AI safeguards

Every model that read this

Model	Provider	Stage	Score	Conf.	Latency	Prompt	When
Llama 3.3 70B	Meta	analysis	-80	80%	4337ms	v1.0.0 / m1.0.1	2026-09-11 12:17
GPT-4.1 mini	OpenAI	consensus	-65	70%	3064ms	v1.0.0 / m1.0.1	2026-09-11 12:19
Claude Sonnet 5	Anthropic	consensus	-45	55%	10517ms	v1.0.0 / m1.0.1	2026-09-11 12:19
Gemini 2.5 Flash	Google	consensus	-100	90%	1863ms	v1.0.0 / m1.0.1	2026-09-11 12:19
Mistral Small 3.1 24B	Mistral AI	consensus	-65	80%	4940ms	v1.0.0 / m1.0.1	2026-09-11 12:19

Llama 3.3 70B · reading

AI used for malicious purposes

evidence: reported horizon: n/a

GPT-4.1 mini · reading

The use of AI by the Houthi group to develop ballistic missiles indicates a material risk as it demonstrates how AI can be exploited for military and potentially destabilizing purposes. Although the missile test reportedly failed, it exposes vulnerabilities in AI safeguards, highlighting risks of misuse and escalation in conflict. Evidence is from a reliable news source, but details are limited.

evidence: reported horizon: near capability 65 societal 70 existential 30 economic 10

Claude Sonnet 5 · reading

A militant group's attempt to use an AI model for missile development, even though it failed, demonstrates a real dual-use security risk and a gap in safeguards. Headline-only sourcing limits detail on causation and severity, so magnitude is capped despite the concerning implication.

evidence: reported horizon: n/a

Gemini 2.5 Flash · reading

The item reports that a non-state actor used Anthropic AI to attempt to develop ballistic missiles. Although the missile test reportedly failed, this incident highlights a severe misuse of AI technology for destructive purposes. It suggests a critical vulnerability in AI safeguards against proliferation.

evidence: reported horizon: n/a

Mistral Small 3.1 24B · reading

The use of AI to develop weapons is strongly negative. The failure of the missile test is a mitigating factor.

evidence: primary horizon: n/a

Evidence extracted

Houthis used Anthropic AI to try to build ballistic missiles
Missile test by group apparently failed

The chain

SOURCE     Financial Times: Artificial Intelligence (tier 1)
   ↓
DOCUMENT   08ef81d8-7125-42d1-81c1-7548281a4aba
           https://ft.com/content/8310cf56-ce60-4e6e-8254-5bb470e9a880
   ↓
EVIDENCE   2 extracted excerpts
   ↓
MODEL RUN  5 runs, methodology 1.0.1
   ↓
SCORE      -71  (Strongly adverse)
   ↓
CONFIDENCE 75%