Safety · Concept
Unmeasured Modality, Search, and Citation Factors in AI Safety Benchmarks
Audit showing ChatGPT UI vs API modality and web search materially change safety-benchmark accuracy, consistency, citations, and abstention.
Forecast horizon
Pushes safety claims to report access modality, search state, and multi-run variance.
Connections
Connections · 4
How this node ties into the rest of the map, and the evidence behind each link.
to · constrains 3/10
Science of AI Evaluation Modality and web-search conditions can reverse apparent safety/performance rankings used in deployment claims.
+3 growth to · applies to 3/10
OpenAI Audit compares ChatGPT chat UI versus OpenAI API with and without web search on safety benchmarks.
+3 growth to · informs 3/10
Science of AI Evaluation UI/API and search confounds show safety claims need modality-aware evaluation science, not single-run API accuracy.
+3 growth from · tracked by 3/10
OpenAI The modality/search audit compares ChatGPT chat UI versus OpenAI API on BBQ and SafetyBench.
+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 →