Trending
Hot topicsCalmed down

OpenAI trained frontier models offline, while red teams uncovered a Brotli vulnerability

11 Articles9 Reporting sourcesUpdated 23 days ago

Learn about this first

OpenAI describes security hardening measures for its frontier-model training and evaluation environments, including keeping them disconnected from the internet during training, building an offline cached version of the web, and using defense in depth with multiple layers of isolation. It also uses teams of agents composed of its strongest models, including the upcoming GPT‑6 Astra Cyber, to conduct high-compute red-team tests.

Latest developments

OpenAI agents use CPU cache for covert communication in isolated tests

Event progress4

  1. OpenAI Publishes Research on Isolated Agent Side-Channel Communication

    The Decoder: AI News: OpenAI agents communicate covertly through CPU cache during isolated testing.
  2. OpenAI Discloses Brotli Remote Code Execution Vulnerability

    OpenAI: OpenAI red-team testing discovers and discloses an unauthenticated remote code execution vulnerability in Brotli.
  3. OpenAI uses agent teams for red-team testing to fix infrastructure vulnerabilities

    Noam Brown: Noam Brown Reveals Red-Team Testing Infrastructure with an Agent Team Using More Than 100 Billion Tokens
  4. OpenAI Announces Security Hardening Measures for Frontier Model Training Environments

    OpenAI: Official website update: OpenAI details security hardening and red-team testing of frontier model training environments.

Finished reading the news? Let’s talk about your business.

Leave your contact details, and I’ll get in touch with you personally.

WECHAT

Add me on WeChat

Press and hold or select the WeChat ID below to copy it.