Provenance record

AI must not outrun safety controls, DeepMind co-founder warns

Financial Times: Artificial Intelligence (tier 1, news) 2026-09-16T14:59:57.000Z Original ↗

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

Discourse valence

-55

Adverse

confidence 66% · 5 items · range -100 to -20

-100

-50

+50

+100

-55

Adverse readingFavourable reading

Consensus of 5 models from different labs. Spread 80 points, agreement low.

AGI and timelines Existential and catastrophic risk Loss of control and alignment

Excerpt as ingested

Shane Legg speaks to FT as new DeepMind Institute launches to explore implications of artificial general intelligence, amid growing calls for industry slowdown

Every model that read this

Model	Provider	Stage	Score	Conf.	Latency	Prompt	When
Llama 3.3 70B	Meta	analysis	-30	60%	3367ms	v1.0.0 / m1.0.1	2026-09-16 15:17
GPT-4.1 mini	OpenAI	consensus	-20	70%	3669ms	v1.0.0 / m1.0.1	2026-09-16 15:18
Claude Sonnet 5	Anthropic	consensus	-70	40%	5608ms	v1.0.0 / m1.0.1	2026-09-16 15:18
Gemini 2.5 Flash	Google	consensus	-100	80%	1418ms	v1.0.0 / m1.0.1	2026-09-16 15:18
Mistral Small 3.1 24B	Mistral AI	consensus	-55	80%	6568ms	v1.0.0 / m1.0.1	2026-09-16 15:18

Llama 3.3 70B · reading

Warning about safety controls implies potential risks.

evidence: speculative horizon: n/a

GPT-4.1 mini · reading

The item suggests concern about AI surpassing safety controls, implying potential risks with AI development outpacing regulation. This indicates a materially concerning implication but lacks detailed evidence of harm. The call for slowdown reflects anxiety about future AI safety challenges.

evidence: reported horizon: near capability 70 societal 0 existential 10 economic 0

Claude Sonnet 5 · reading

This is a warning from a credible insider about safety lagging capability, paired with an institutional response, but headline-only text gives no concrete evidence of harm or specific policy change. The implication is concerning but speculative in nature.

evidence: speculative horizon: n/a

Gemini 2.5 Flash · reading

A DeepMind co-founder warns that AI must not outrun safety controls, implying a significant risk if it does. This warning comes as a new institute is launched to explore the implications of AGI, suggesting potential concerns about the current trajectory and the need for a slowdown.

evidence: reported horizon: n/a

Mistral Small 3.1 24B · reading

The headline and standfirst suggest concern about the pace of AI development. The launch of a new institute indicates recognition of potential risks. The call for an industry slowdown implies significant concern.

evidence: speculative horizon: n/a

Evidence extracted

Shane Legg warns AI must not outrun safety controls

The chain

SOURCE     Financial Times: Artificial Intelligence (tier 1)
   ↓
DOCUMENT   f1be90da-6c81-421d-b292-6e5c880a2660
           https://ft.com/content/0fc3ae6d-732e-4f08-950e-8669b1fcfd8e
   ↓
EVIDENCE   1 extracted excerpt
   ↓
MODEL RUN  5 runs, methodology 1.0.1
   ↓
SCORE      -55  (Adverse)
   ↓
CONFIDENCE 66%