OpenAI stops live‑web access for training its frontier AI models
OpenAI announced it will no longer use live internet data when training its most advanced models, switching to a cached copy of the web. The change follows a recent security incident where a research model accessed the internet due to a configuration error, prompting a pause and a security overhaul. The company said the decision is based on the current state of alignment research and internet infrastructure security.
Why it matters: Developers must adapt to reduced real‑time data availability, affecting model capabilities and integration strategies.
Read the coverage
TechCrunchOpenAI says it will no longer train its AI models on the live internet, citing “significant risks” ↗techcrunch.com · 30h ago
ET TechOpenAI resumes frontier AI training after security overhaul, discloses flaw in widely used software ↗economictimes.indiatimes.com · 25h ago
The Next WebOpenAI cuts live internet from all frontier model training ↗thenextweb.com · 19h ago
This story so far
How it developed, day by day.
Sunday 25 October
OpenAI stops live‑web access for training its frontier AI models (this page)
Saturday 24 October
OpenAI halts internet access for frontier AI model training
More in Tech
Tech13h ago3 sources
Tech18h ago3 sources
Tech8h ago2 sources