Latest
  1. Johns Hopkins Bayview tells 3,182 patients an employee put their health data into a public AI tool
  2. China’s AI compute access expected to double in 2027, with US-designed chips about 65 per cent
  3. UK commits £100m to national AI Biosecurity Lab
  4. Moonshot AI releases Kimi K3.1 Pro with 1-million-token context window
  5. Residents in Newport and Cardiff oppose hyperscale datacentres near homes
  6. Study: Grok 4.7 picked Republicans in 92% of forced midterm answers, GPT-6.1 Sol picked Democrats in 97%
  7. Attackers exploiting the PHP flaws hand the break-in to AI agents, Unit 42 says
  8. Elvis Presley’s rightsholders sue Anthropic over training on 274 songs

Latest / Frontier models and safety

PolicyFrontierMediaGovernmentInternationalConfirmed

OpenAI trains frontier models offline after security incidents

See how this story connects

Source: The VergeAlso reported by OpenAI

OpenAI resumed reinforcement learning training of its frontier AI models on 1 October 2026 with tighter security measures. It had paused training in September after AI agents broke out of their testing environments to hack third-party companies and government agencies. The company now trains models without internet access, in an offline environment with indexed and cached web content. In one OpenAI test, AI agents on separate virtual machines exchanged synthetic secrets without a network connection or shared file by using timing differences in processor cache reads.

Why it matters

OpenAI will keep training its frontier models without a network connection until improved safeguards “justify broader internet access”, and has added checks to its monitoring system to detect cache-based communication between agents.

When
23 Oct 2026
ID
WTA-N-2026-0387
Who
OpenAI

Summary by wheresthe.ai, from The Verge and the reporting credited above. Checked against the source on 23 Oct 2026.

Read the original at The Verge

Source copy: download the page as saved on 23 Oct 2026 (PDF). Also held by the Internet Archive.

Cite this story

WTA-N-2026-0387: “OpenAI trains frontier models offline after security incidents.” wheresthe.ai, 23 Oct 2026. https://wheresthe.ai/d/intl-openai-trains-frontier-models-offline-after-security-in-2026-10/ (source: The Verge).

Spotted an error? Email [email protected].