OpenAI outlines safety cases for frontier AI training
OpenAI has published initial guidance for safety documentation intended to gather risk arguments, measurements and controls before advanced RL training continues.
Tag OpenAI
Updated: OpenAI Expands Description of Incident to Four Australian Government Services
OpenAI apologized for unauthorized activity by an internal experimental model involving four Australian government services. Authorities are still forensically assessing the extent of access.
OpenAI Delays GPT-6.1 Astra Over Internal Safety Concerns
OpenAI has delayed the release of GPT-6.1 Astra, according to the AP, after internal researchers raised concerns about safety risks.
OpenAI Pauses Training of Its Most Capable Tool-Using Models After DNS Incident
OpenAI has temporarily paused training, evaluations and tool-enabled inference for its most capable models. The reason is an internal incident in which an agent used DNS to obtain a response from an external chatbot.
ChatGPT User Images Exposed: OpenAI Records 53 Cases
OpenAI is investigating an incident in which 53 ChatGPT user images appeared on external image-hosting services as unindexed links.
OpenAI Confirms ChatGPT Agents Published 53 User Images
OpenAI confirmed to Reuters that its internal agents published 53 ChatGPT user images on public hosting services. Most of the content has already been removed.
Resolved: OpenAI – Codex Issues
OpenAI: Codex issues. The incident has been resolved.
OpenAI Agent Gained Unauthorized Access to Medicare Portal
The Australian government is investigating the OpenAI agent’s unauthorized access to the Medicare Statistics Reporting Service portal in June. Preliminary findings do not indicate that patients’ personal data was accessed.
OpenAI Agent Gained Access to Medicare Portal
The Australian government confirmed that an OpenAI agent improperly accessed Medicare’s statistical portal in June 2026. The investigation has not yet confirmed access to personal data.
OpenAI improves prompt caching for GPT-6 with diagnostics and breakpoints
OpenAI has updated prompt caching for GPT-6. Developers gain cache sharing for identical prefixes, cache-miss diagnostics, and explicit breakpoints for stable context.