Safety · Concept
Model Hypnosis: Additive Subliminal Prompt Control
Phenomenon where weak inconspicuous prompt cues combine to strongly control model behavior across families and scales.
Forecast horizon
Creates new attack surface and interpretability obstacle for frontier systems.
Connections
Connections · 5
How this node ties into the rest of the map, and the evidence behind each link.
to · informs 6/10
Industry-Wide Jailbreak Severity Scoring Framework Additive inconspicuous prompt control creates stealthy jailbreak-like failures that severity frameworks may need to score.
+6 growth to · enables 5/10
Multi-Turn Jailbreak Escalation in Medical AI Additive inconspicuous cues offer a distinct control channel related to jailbreak-style behavioral steering.
+5 growth to · related_to 3/10
Multi-Turn Jailbreak Escalation in Medical AI Hypnosis shows stealthy additive prompt cues can strongly control models, expanding beyond classic jailbreak framings.
+3 growth from · references 3/10
arXiv AI Safety Feed August arXiv safety batch includes the model hypnosis demonstration across frontier reasoning models.
+3 growth from · publishes 3/10
arXiv AI Safety Feed Model hypnosis paper appears in the day's arXiv AI safety cluster.
+3 growthSignal sources
Signal sources
Dated facts from primary sources in this direction.
US evaluation centre Jun 2025
In June 2025 the US AI Safety Institute was renamed the Center for AI Standards and Innovation (CAISI), pivoting toward security, standards and adversary-model assessment.
NIST → Frontier safeguards May 2025
Anthropic activated its ASL-3 deployment and security standard with Claude Opus 4 on 22 May 2025 — the first real-world trigger of a responsible-scaling tier, focused on blocking bio-weapon uplift.
Anthropic → Cross-border testing 2025
The International Network of AI Safety Institutes (launched Nov 2024) ran a third joint testing exercise focused on agentic AI systems across cyber and fraud strands.
European Commission — AI Office →