Model Hypnosis: Strong control of AI via additive subliminal effects
We demonstrate that AI models are broadly susceptible to a phenomenon we call model hypnosis, in which individually weak and seemingly irrelevant cues in the prompt can be systematically combined to strongly control model behavior. Model hypnosis occurs across model families and scales, including in frontier reasoning models, and hypnotic prompts can transfer between models. Because the model is controlled by inconspicuous textual choices, such as paraphrases and typos, model hypnosis presents new challenges and avenues for AI safety, and is a major hurdle for AI interpretability.
Lineage graph
Paper → model → repo connections mined from source citations (Tier-1 exact match).
Why these links exist
Every edge carries a method, confidence, and the source snippet that justified it — so bad links are debuggable.
- FuzzySimilar title/name (fuzzy) · 87%lllyasviel/ControlNet-v1-1 →
“Fuzzy title match (0.94): “Model Hypnosis: Strong control of AI via additive subliminal” ≈ “lllyasviel/ControlNet-v1-1””
- FuzzySimilar title/name (fuzzy) · 87%lllyasviel/ControlNet →
“Fuzzy title match (0.94): “Model Hypnosis: Strong control of AI via additive subliminal” ≈ “lllyasviel/ControlNet””
- PossiblePossibly related (embedding) · 52%Escape Artists: 'Incorrigible' AI Models Resist Rehabilitation - Dark Reading →
- PossiblePossibly related (embedding) · 47%AI Hallucinations: Why Artificial Intelligence Can Sound Convincing While Getting the Facts Wrong - USA Herald →
- PossiblePossibly related (embedding) · 47%Experts Say There’s Now an Open Source AI Model as Scary as Mythos - Futurism →
- PossiblePossibly related (embedding) · 46%Chain-of-Thought Spoofing Targets Reasoning AI Models - Hackaday →
- PossiblePossibly related (embedding) · 45%The ‘dangerous’ AI models are the ones saving us - washingtonexaminer.com →
- FuzzySimilar title/name (fuzzy) · 59%builderz-labs/mission-control →
“Fuzzy title match (0.73): “Model Hypnosis: Strong control of AI via additive subliminal” ≈ “builderz-labs/mission-control””