Safety · Concept
Language-Specific Gaps in Multilingual AI Safety Datasets
Audit shows multilingual safety coverage claims fail per-language, with total gaps in self-harm/sexual categories for African languages studied.
Forecast horizon
Providers will face pressure for per-language safety dataset SLAs, not collection-level coverage claims.
Connections
Connections · 3
How this node ties into the rest of the map, and the evidence behind each link.
to · constrains 3/10
Science of AI Evaluation Language-slice gaps show collection-level multilingual safety benchmarks can overstate per-language protection.
+3 growth to · constrains 3/10
Rules vs Character Scaling Laws for AI Safety Design Per-language safety data holes limit both character-shaping and rule-filter effectiveness for non-English users.
+3 growth to · informs 3/10
Safety Benchmark Transfer Gap for Small Language Models Both show safety evaluation claims fail when transferred across model classes or language resource tiers.
+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 →