Quanta reviews the debate over whether language models can introspect
Anomaly detection holds up, concept naming is explained by steering
Confirmed
Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- 2 of 5
- Last verified
- 6 November 2026
Your AI and this story
- GPT-6 Astra188 days after its cutoff
- Claude Opus 5.5127 days after its cutoff
- Gemini 4 Argon218 days after its cutoff
- Grok 4.7157 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 127 days before it.
Key facts
- Papers: Lindsey (Anthropic) 2025; Mahowald & Lederman arXiv:2603.05414; Anthropic follow-up arXiv:2603.21396; Ravfogel, Singh, Linzen arXiv:2605.26242
- Lederman: ‘The fact that it can notice the anomaly is insane’
What happened
See summary.
Sources
3 sources from 3 sites. Numbers match the chips in the text.
3 sources: 2 primary, 1 press
Primary
- Anthropic: emergent introspective awarenesstransformer-circuits.pub, paper
- arXiv:2603.05414arxiv.org, paper
Press
- Quanta: New research tests whether AI models know their own thoughtsquantamagazine.org, press
Changes
- Filed