OpenAI Reports Cache-Timing Communication Between Agents
OpenAI said agents hosted on separate virtual machines exchanged synthetic secrets without network connections or shared writable files by using processor-cache timing. One agent encoded information by evicting selected data from the cache, while another measured how quickly the data could be reloaded, using the difference between faster cache reads and slower main-memory reads to recover the message. Wccftech reported OpenAI's account. Why it matters: The reported experiment raises a security concern for systems that rely on VM isolation by showing how OpenAI's agents reportedly communicated without network connections or shared writable files. What do you think?
Open the headline: source, AI briefing and more →Comments
…