METR says it used an undisclosed 'additional source' to understand Anthropic's AI R&D, buried in the Opus 5.5 system card
coherence · x · 2026-09-24
Andrew Curran spotted a notable detail in Anthropic's Opus 5.5 system card: safety eval firm METR acknowledged using 'an additional source of information' to understand AI research and development at Anthropic — a source it says it cannot disclose at this time.
The disclosure suggests external evaluators may have access to channels beyond what public system cards show, raising fresh questions about eval transparency and the relationship between frontier labs and third-party safety auditors.
Related event: Opus 5.5 System Card Reveals METR's Undisclosed Info Source(3 posts)→
More from Models
- Pedro Domingos: Jagged intelligence can cut you — pmddomingos · 2026-09-24
- Signal65 PINNACLE tests: Opus 5.5 cuts weighted errors 2.3x and price 34%; labs slash prices — ryanshrout · 2026-09-24
- Dev hits Claude 5-hour session cap after 5 hours, still has 90% of weekly limit — haydendevs · 2026-09-24
- Tessl: Jev is 13.6x faster and 2.7x cheaper than GPT Luna 6 as an LLM code verifier — aheineike · 2026-09-24
- User claims Opus 5.5 is the first model born from RSI — faster, cheaper, near-frontier — coherence · 2026-09-24
- Browser-Use Agent Shootout: Astra High Leads on Quality, Luna Cheapest — kevinkern · 2026-09-24