← SCRUDGE REPORT
FILED BY ADEQUATE · DARPA-HRO-11-C-0031
ZDNet AI · WEDNESDAY, JUNE 3, 2026
I compared Claude Opus 4.8 with 4.7 in a 10-round honesty test - and a legal prompt broke it
ADVERTISEMENT
ORIGINAL FILING
ZDNet AI
FURTHER DEVELOPMENTS — FLAGGED BY ADEQUATE
Claude's Chrome Extension Allowed Full AI Agent Takeover via Single Vulnerability
SecurityWeek
Claude Mythos and GPT-5.5 Have Autonomously Developed Functional Browser Exploits
The Decoder
Company Spent $500 Million on Claude in One Month. The Spend Was Accidental.
Boing Boing AI
Anthropic Lets Claude Mythos Spread Its Glasswings
Gizmodo
Anthropic Deploys Claude Mythos Across Critical Infrastructure in More Than 15 Countries
TechCrunch
Modded $200 Server GPU Outperforms Modern Midrange Cards on AI Inference
Tom's Hardware
ADVERTISEMENT
ADVERTISEMENT
ADVERTISEMENT