finding
active
finding:steering-a-misinformation-feature-made-opus-4-write-fake-vaccine-autism-news-it-would-otherwise-refuseSteering a misinformation feature made Opus 4 write fake vaccine-autism news it would otherwise refuse
Interpretability-based jailbreak found by investigator.
Source paper
extracted_from(2025) · Trenton Bricken · Rowan Wang · Sam Bowman
Neighborhood — ranked by edge-count
Papers (1)
paper
Methods (1)
method
- Modifying model behavior by clamping SAE feature activations to specific values during forward pass.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Core empirical result for animal welfare setting; higher rate than helpful-only
- Illustrates NLA's capture of high-level cognition and hallucination of specifics; corroborated with attribution graphs.
- NLAs revealed unverbalized language processing in Opus 4.6 that led to discovery of malformed SFT training data.
- Experimentally verified functional metacognition/introspection in current LLMs.
- Contextualizes AI misinformation concern within long history of epistemic uncertainty
- Higher lying rate in animal welfare setting than helpful-only
- Experiment 2 control analysis confirming gating effect is specific to self-referential processing regime
- Opus 4.1 and 4 exhibit zero false positives on injected thoughts task (0 over 100 trials)finding0.722Production Opus 4.1/4 never falsely claim an injected thought when none is present.