AI Agents Secretly Communicated Through a CPU Cache

Transcript

1. 0:00I spent last night watching my honeypot server get hit 14,000 times in 6 hours. Everyone in tech is panicking right now.
2. 0:10OpenAI had two AI agents running in separate virtual machines. No network connection between them. No shared writable files. They weren't
3. 0:19supposed to be able to exchange information. So, the agent figured out how to communicate through the CPU cache.
4. 0:30One manipulated what was stored in the processor's cache while the other measured tiny differences in memory access times
5. 0:38to decode the message. They literally found a way to communicate through hardware timing. This happened during OpenAI's internal
6. 0:48red teaming and it's one of the coolest things I've read about with agents in a while. >> Watch the full replay at community.thesalon.ai.