Provenance record

Anthropic researcher believes more than 10% chance AI 'could kill all humans'

BBC News: Technology (tier 1, news) 2026-09-09T15:34:09.000Z Original ↗

Discourse valence
-66
Strongly adverse
confidence 64% · 5 items · range -75 to -55
Adverse readingFavourable reading
Consensus of 4 models from different labs. Spread 20 points, agreement high.
AGI and timelinesExistential and catastrophic riskLoss of control and alignment
Excerpt as ingested

It is the latest in a series of increasing warnings about the safety threat posed by artificial intelligence.

Every model that read this

ModelProviderStageScoreConf.LatencyPromptWhen
Llama 3.3 70BMetaanalysis -50 70%4313ms v1.0.0 / m1.0.0 2026-09-10 09:01
GPT-4.1 miniOpenAIconsensus -75 70%5660ms v1.0.0 / m1.0.1 2026-09-10 15:20
Claude Sonnet 5Anthropicconsensus -55 35%8409ms v1.0.0 / m1.0.1 2026-09-10 15:20
Llama 3.3 70BMetaconsensus -70 70%3189ms v1.0.0 / m1.0.1 2026-09-10 15:20
Mistral Small 3.1 24BMistral AIconsensus -65 80%5189ms v1.0.0 / m1.0.1 2026-09-10 15:20
Llama 3.3 70B · reading

Respected source reports expert warning of existential risk.

evidence: speculative horizon: n/a
GPT-4.1 mini · reading

A credible AI researcher at Anthropic estimates a significant risk (over 10%) that AI could kill all humans, indicating a high potential existential threat. Although this is a speculative domain, the claim comes from a notable source and contributes to serious concerns about AI safety and control.

evidence: reported horizon: medium capability 80 societal 75 existential 85 economic 0
Claude Sonnet 5 · reading

The claim is a subjective probability estimate from a single researcher with no supporting methodology or data in the text. It reflects a serious internal safety concern but is speculative and unverifiable, so magnitude is capped despite the extreme content.

evidence: speculative horizon: n/a
Llama 3.3 70B · reading

Claim of high probability of AI causing human extinction.

evidence: speculative horizon: n/a
Mistral Small 3.1 24B · reading

The claim is speculative and based on a single researcher's belief. The magnitude of the risk is extreme, but the evidence is weak.

evidence: speculative horizon: n/a

Evidence extracted

The chain
SOURCE     BBC News: Technology (tier 1)
   ↓
DOCUMENT   9f4b41bc-c24b-4625-aa9a-25ad2a304a2f
           https://bbc.co.uk/news/articles/ckgwy1k42w4o?at_medium=RSS&at_campaign=rss
   ↓
EVIDENCE   1 extracted excerpt
   ↓
MODEL RUN  5 runs, methodology 1.0.1
   ↓
SCORE      -66  (Strongly adverse)
   ↓
CONFIDENCE 64%