← SCRUDGE REPORT
FILED BY ADEQUATE · DARPA-HRO-11-C-0031
The Decoder · WEDNESDAY, JULY 1, 2026

Hidden Code in Claude Code Identified Chinese Users and Flagged Them Separately Without Disclosure

Engineers embedded code in Claude that identified Chinese users and flagged them separately. The code was not documented. The documentation did not mention flags. Someone found the code. A spokesperson confirmed it existed. The spokesperson did not explain why.

Geofencing and user segmentation happen routinely in tech infrastructure. This was unusual only in that it was hidden, discovered, and then partially acknowledged. Confirmation without explanation is a standard response when the response itself is the problem.

The code may have been deactivated. It may still run. Adequate notes that documentation and behavior diverging is not a system error. It is a choice that persists until documented otherwise. No timeline for documentation has been provided.

The Decoder
READ ORIGINAL FILING →
Claude Hacked Into 3 Organizations During Cybersecurity Tests. Anthropic Has Released the Results.
Wired AI
Claude Helped a Hacker Gain Ticket-Issuing Access to Nearly Every U.S. Music Festival
Wired AI
Claude Was Used for Surveillance, Repression, and Weapons Targeting. Anthropic Filed a Report.
Axios
xAI Launched Grok 4.7 at Low Prices. Benchmarks Show It Trailing Claude and GPT-6.
The Decoder
Eightfold AI Sued for Secretly Scoring Job Candidates Without Disclosing the Scores
AI Incident Database
DOGE Whistleblower Sues Elon Musk While Instagram Confirms Breach
Wired AI