Safety · Concept
Science of AI Evaluation
Joint Singapore-Korea AI Safety Institute evaluation of agent data leakage across 12 realistic non-adversarial tasks demonstrates no agent achieved fully safe execution.
Forecast horizon
Connections
Connections · 7
How this node ties into the rest of the map, and the evidence behind each link.
to · informs 4/10
AISI Frontier AI Trends Report AISI's Frontier AI Trends Report exemplifies evidence-based AI evaluation science applied to frontier model assessment.
+4 growth from · informs 4/10
LLM Reasoning as Pattern Matching Understanding that LLM reasoning is pattern-matching rather than abstract world modeling informs how AI evaluation science should be designed.
+4 growth from · supports 3/10
Ada Lovelace Institute Ada Lovelace Institute published commentary on strengthening the science of AI evaluation to bring clarity to AI risks and benefits.
+3 growth from · supports 3/10
Systems-Safety Methods for Agentic AI Loss-of-Control Risk Systems-safety methods applied to agentic AI strengthen the science of AI evaluation by surfacing risks missed by model-level testing.
+3 growth from · supports 3/10
Reward Hacking in Text-Based AI Safety Gridworlds Text-based AI Safety Gridworlds provide controlled evaluation infrastructure for studying reward hacking, strengthening the science of AI evaluation.
+3 growth from · supports 3/10
NRT-Bench: Multi-Turn Red-Teaming Benchmark for Safety-Critical AI Agents NRT-Bench advances the science of AI evaluation by providing objective harm signals rather than LLM-judged text for safety-critical agent assessment.
+3 growth from · supports 3/10
Industry-Wide Jailbreak Severity Scoring Framework Industry-wide jailbreak severity scoring standardizes evaluation of adversarial attacks across AI systems.
+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 →