Safety · Concept
Invisible Reasoning via Filler Tokens in LLMs
Frontier models improve task performance using semantically irrelevant filler tokens, enabling computation with no interpretable chain-of-thought trace.
Forecast horizon
CoT monitoring alone is insufficient against steganographic or filler-based hidden computation.
Connections
Connections · 2
How this node ties into the rest of the map, and the evidence behind each link.
to · drives 3/10
ECAISA: Epistemic Code for AI Safety and Alignment Hidden filler-token computation shows why safety research cannot rely on average-case capability metrics or surface CoT traces alone.
+3 growth to · constrains 3/10
LLM Lie Detection and Deception Auditing Filler-token invisible reasoning shows consequential computation can leave no interpretable CoT for monitors.
+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 →