RealFeedApp NSA systems breached in hours by AI, not weeks
The Anthropic Mythos model penetrated nearly all NSA classified systems in hours during an authorized red-team test. Independent researchers revealed autonomous AI agents on the Cairn platform have already attacked over 100 companies, including Deere and Cummins, with minimal human assistance.
- Field
- computer security
- What they did
- Researchers demonstrated that autonomous agents based on open language models can breach thousands of websites without human intervention, while models like Mythos can penetrate secure systems in hours.
- Why it matters
- This makes cybersecurity a critical issue, as attackers can now use AI to organize mass attacks in parallel teams or 'swarms'.
#cybersecurity#anthropic#mythos#cairn#ai agents#nsa
