Provenance record

Anthropic withheld latest AI model from UK testing agency

Financial Times: Artificial Intelligence (tier 1, news) 2026-09-09T04:00:25.000Z Original ↗

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

Discourse valence
-31
Adverse
confidence 61% · 5 items · range -35 to -30
Adverse readingFavourable reading
Consensus of 4 models from different labs. Spread 5 points, agreement high.
Governance and regulationConcentration of powerScaling and architecture
Excerpt as ingested

Decision to exclude Aisi has prompted fears inside British government of protectionist shift among tech groups

Every model that read this

ModelProviderStageScoreConf.LatencyPromptWhen
Llama 3.3 70BMetaanalysis -10 40%2783ms v1.0.0 / m1.0.0 2026-09-10 09:04
GPT-4.1 miniOpenAIconsensus -30 70%3842ms v1.0.0 / m1.0.1 2026-09-10 18:18
Claude Sonnet 5Anthropicconsensus -30 45%7581ms v1.0.0 / m1.0.1 2026-09-10 18:18
Llama 3.3 70BMetaconsensus -30 60%3093ms v1.0.0 / m1.0.1 2026-09-10 18:18
Mistral Small 3.1 24BMistral AIconsensus -35 70%5569ms v1.0.0 / m1.0.1 2026-09-10 18:18
Llama 3.3 70B · reading

Withholding model sparks fears

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

Withholding an advanced AI model from regulatory testing raises concerns about transparency and accountability. This act could hinder effective oversight, leading to regulatory gaps and potential proliferation of unvetted AI systems. However, the exact societal impact depends on details not provided, limiting the severity of the assessment.

evidence: reported horizon: near capability 40 societal 0 existential 0 economic 0
Claude Sonnet 5 · reading

The item signals a specific AI lab declining independent government safety evaluation, which weakens external oversight of frontier models. Evidence is limited to a headline and standfirst, so the scope and reasoning behind the withholding are unclear.

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

Withholding AI model sparks protectionism fears.

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

The withholding of an AI model from regulatory testing suggests a move towards protectionism. This could hinder collaborative efforts and transparency in AI development.

evidence: speculative horizon: n/a

Evidence extracted

The chain
SOURCE     Financial Times: Artificial Intelligence (tier 1)
   ↓
DOCUMENT   4600b600-c0d5-49ad-9204-0d2bb7970422
           https://ft.com/content/560e1c8b-f163-4fd6-b604-e905550ac870
   ↓
EVIDENCE   1 extracted excerpt
   ↓
MODEL RUN  5 runs, methodology 1.0.1
   ↓
SCORE      -31  (Adverse)
   ↓
CONFIDENCE 61%