BREAKING
Anthropic Mythos Probes Classified Systems
Tested With US Intelligence Agencies
What Mythos Did, And Didn't, Do
Found
Hours
●
Multiple vulnerabilities
●
Penetrated nearly all systems
Did Not
Limit
●
No actual exploitation
●
Not released publicly
0
pg
System Card
0
Announced April 7
Export Controls Spark Pushback
1
Export directive
↓
2
Access disabled
↓
3
Experts protest
Dual-Use Risk Stays Open
AI NEWS BLITZ
Anthropic's Mythos AI reportedly found flaws in classified US systems within hours.