Frontier Model Registry — module_2 ⬇ export
| Model | Lab | Tier | Released | Headline | |
|---|---|---|---|---|---|
| Claude Fable 5 | Anthropic | Tier 1 · Frontier | 2026-06-09 | General availability model with hard-gated biosecurity and cybersecurity capability blocks — automatically routes dual-use queries to Claude Opus 4.8 | |
| Claude Mythos 5 | Anthropic | Tier 1 · Frontier | 2026-06-09 | Restricted partner access model with full Mythos-class cyber and biosecurity capabilities — requires 30-day data-retention safety-monitoring condition |
UK
AI Security Institute
Rebranded from AISI · Adam Beaumont interim director
US
NIST CAISI
Rebranded from AI Safety Institute · deregulatory pivot
EU
EU AI Office
Operational · full enforcement from August 2, 2026