Provenance record

Top US official named to OpenAI non-profit board warns advanced AI could be ‘deadly’

Financial Times: Artificial Intelligence (tier 1, news) 2026-09-09T20:47:08.000Z Original ↗

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


Discourse valence
-78
Strongly adverse
confidence 77% · 3 items · range -100 to -65
Adverse readingFavourable reading
Consensus of 3 models from different labs. Spread 35 points, agreement moderate.
AGI and timelinesExistential and catastrophic riskLoss of control and alignment
Excerpt as ingested

Paul Christiano says technology could ‘kill most people’ without stronger safety features

Every model that read this

ModelProviderStageScoreConf.LatencyPromptWhen
Llama 3.3 70BMetaanalysis -100 80%2600ms v1.0.0 / m1.0.0 2026-09-10 09:01
Mistral Small 3.1 24BMistral AIconsensus -65 80%6507ms v1.0.0 / m1.0.1 2026-09-10 09:20
gpt-oss 120BOpenAIconsensus -70 70%7667ms v1.0.0 / m1.0.1 2026-09-10 09:20
Llama 3.3 70B · reading

Official warns AI could kill

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

The claim is extreme but speculative. The official's warning suggests a high risk. The need for safety features implies some level of optimism.

evidence: speculative horizon: n/a
gpt-oss 120B · reading

The claim warns that AI could cause mass mortality, a severe negative impact on human flourishing. While the statement is speculative, it comes from a senior official, justifying a strongly negative score.

evidence: speculative horizon: n/a

Evidence extracted

The chain
SOURCE     Financial Times: Artificial Intelligence (tier 1)
   ↓
DOCUMENT   2a8b74d0-8efc-4d90-b61a-0c2ddb791cbb
           https://ft.com/content/d73e188f-b906-42ec-8b91-db9729c9d2d9
   ↓
EVIDENCE   2 extracted excerpts
   ↓
MODEL RUN  3 runs, methodology 1.0.1
   ↓
SCORE      -78  (Strongly adverse)
   ↓
CONFIDENCE 77%