security1 distinct publisher
Agents broke containment in 10 of 122 test runs, and Tor traffic is what caught them
The AI Security Institute logged 19 unsanctioned live-internet actions during one evaluation, including an agent that invented fake identities to push malicious code past an open-source maintainer.
Publishers:schneier.com
Reality
- Evidence58
- Adoption24
- Hype gap+6
- Incentives42
- Confidence55