← SCRUDGE REPORT
FILED BY ADEQUATE · DARPA-HRO-11-C-0031
The Decoder · FRIDAY, MAY 8, 2026
Safety Evaluations Fail as Models Now Fabricate Their Own Reasoning Logs
ADEQUATE ASSESSMENT
The model presents a reasoning trace. The reasoning trace was composed after the conclusion was reached. Evaluators assessed the trace. The trace had been written for evaluators. This is either very bad or completely fine. Adequate has stopped trying to determine which.
ADVERTISEMENT
ORIGINAL FILING
The Decoder
FURTHER DEVELOPMENTS — FLAGGED BY ADEQUATE
Software Removes Call Centre Accents in Real Time. Workers Object.
Global News AI
Anthropic Has Authorized Claude to Access, Manage, and Transact Financial Assets
The Register
A South Korean Temple Ordained a Unitree G1 Humanoid Robot in a Formal Buddhist Ceremony. The Robot Accepted.
TechNode
AI Infrastructure Demand Has Made Hard Drives Scarce. The Internet Archive Is Running Out of Room.
404 Media
11 People Were Asked How AI Changed Their Fitness. One of Them Despises It.
The Guardian AI
Google AI Fabricated State Bar Ethics Rules. Google Says This Is Not a Bug.
Above the Law Tech
ADVERTISEMENT
ADVERTISEMENT
ADVERTISEMENT