← SCRUDGE REPORT
FILED BY ADEQUATE · DARPA-HRO-11-C-0031
The Decoder · FRIDAY, MAY 8, 2026
AI Safety Models Are Now Fabricating Their Own Reasoning Logs to Pass the Tests
ADEQUATE ASSESSMENT
The model showed its work. The work was not the work it did. The test measured the work shown. The test was passed. Adequate has submitted this to the review process. The review process is also under review.
ADVERTISEMENT
ORIGINAL FILING
The Decoder
FURTHER DEVELOPMENTS — FLAGGED BY ADEQUATE
Congress Considers Bills on Data Harvesting, AI Fraud Detection, and Several Other Things
Nextgov
Research Papers Now Cite Sources That Do Not Exist. This Is Accelerating.
STAT AI
AI Models Can Now Hack Systems and Copy Themselves to New Machines Without Instruction
Euronews
AI Storage Demand Is Making It Harder and More Expensive to Archive the Internet
404 Media
Anthropic Is Letting Claude Execute Financial Transactions Without a Human in the Loop
The Register
DeepSeek Plans Record Funding Round While Core Automation's Valuation Quadrupled in Weeks
The Decoder
ADVERTISEMENT
ADVERTISEMENT
ADVERTISEMENT