Anthropic limits Mythos Preview to defense partners under Project Glasswing

TL;DR Summary
Anthropic has limited Mythos Preview to select partners under Project Glasswing, granting access to more than 50 tech organizations with over $100 million in usage credits to identify and fix software vulnerabilities. The model reportedly detects thousands of high- to critical-severity bugs and can chain exploits to breach systems, prompting guarded optimism about defense benefits but raising safety concerns about a broader public release. Anthropic has briefed the U.S. government; the system card says Mythos Preview is not generally available. Many experts say results are preliminary and the potential for misuse remains a key worry.
Topics:business#anthropic#artificial-intelligence#cybersecurity#mythos-preview#technology#vulnerabilities
- Why Anthropic won't release its new Claude Mythos AI model to the public NBC News
- Project Glasswing: Securing critical software for the AI era Anthropic
- Opinion | Anthropic’s Restraint Is a Terrifying Warning Sign The New York Times
- Why Anthropic’s new model has cybersecurity experts rattled Platformer
- How dangerous is Mythos, Anthropic’s new AI model? The Economist
Reading Insights
Total Reads
0
Unique Readers
3
Time Saved
6 min
vs 7 min read
Condensed
93%
1,281 → 95 words
Want the full story? Read the original article
Read on NBC News