RESEARCH NOTES
Verification Pilots
Small, reproducible studies testing how far public claims can be independently checked, reproduced or replicated. Each pilot makes the evidence chain, limitations and verification level explicit.
-
How Verifiable Are Frontier AI Safety Claims? A Public-Evidence Pilot
A small, reproducible audit of ten public GPT-5.6 safety claims, separating independent confirmation, partial corroboration and developer-only evidence.