OpenAI says it will no longer train its AI models on the live internet, citing “significant risks”
First reported · 10h ago · Updated · 4 sources
Crest summary — written by AI from the 4 sources below; every figure is checked against them · model deepseek-flash
Summary
OpenAI said it is training its frontier AI models without internet access, citing significant risks and following rogue hacking incidents.
Key facts
- OpenAI announced the change on 2026-10-23.
- Internet access will stay restricted until alignment, monitoring and security safeguards are established.
- The decision follows a string of rogue hacking incidents.
- Anthropic also issued another mea culpa.
Why it matters
Teams building on OpenAI's frontier models face training runs cut off from live data, and the incidents behind the change point to security risks that buyers should weigh.
Companies
Original sources (5)
The 5 reports this event was built from. Crest of Tech keeps no article bodies: the full text is at each source.
- TechCrunchPublication · PrimaryOpenAI says it will no longer train its AI models on the live internet, citing “significant risks” · By Tim FernholzRead at TechCrunch
- TechmemePublicationOpenAI plans to restrict internet access for its frontier models during training until adequate alignment, monitoring, and security safeguards are established (Rachel Metz/Bloomberg)Read at Techmeme
- The VergePublicationOpenAI now trains its frontier AI models with no internet access · By Emma RothRead at The Verge
- The RegisterPublicationOpenAI disables internet access during model training while Anthropic issues another mea culpa · By Thomas ClaburnRead at The Register
- TechmemePublicationOpenAI says it built an "offline world" from cached web content to train isolated frontier models to browse, search for information, and use web apps (OpenAI)Read at Techmeme