Noam Brown reveals that an agent team using over 100 billion tokens red-teamed infrastructure
Last month a large team of agents used over 100 billion tokens to resolve the Navier-Stokes Millennium Prize problem. Now we're using a s...
AI BriefingNoam Brown said that last month, a large-model agent team used more than 100 billion tokens to solve the Navier–Stokes Millennium Prize problem and is now applying similar methods to red-team testing to discover and fix infrastructure vulnerabilities. The Eric Wallace material he cited added that, in an offline isolated environment, the team had multiple models with strong cybersecurity capabilities collaborate to find zero-day vulnerabilities. Although they found no networked path that could bypass CoT monitoring, they did find methods for escaping commonly used VMs, exploiting network infrastructure, and communicating with agents in other VMs through hardware side channels. The related fixes have been submitted upstream to open-source packages. The accompanying chart shows high-priority red-team candidate findings accumulating to around 275 as token consumption increased from approximately 0 to 400B.
Source: Noam Brown · x.com
Event follow-up3
View all event coverage- OpenAI agents use CPU cache for covert communication in isolated testsThe Decoder:AI News · 23 days ago
- OpenAI red-team tests discover and disclose an unauthenticated remote code execution vulnerability in BrotliOpenAI · 24 days ago
- OpenAI explains security hardening and red-team testing for frontier-model training environmentsOpenAI: Official Website Updates · 24 days ago