Executive summary
OpenAI Unveils Safety Case Guidelines for Frontier AI Training
OpenAI has published early guidelines for safety cases in frontier AI training (opens in a new tab), outlining technical safeguards and operational practices for high-risk models. This comes alongside an apology to Australia for incidents involving government websites, where the company detailed stronger safeguards, and news of a GPT-6 Astra (opens in a new tab) deployment completing a complex 50-tab tax workbook at double the speed of its predecessor.
Meanwhile, the NCSC is urging UK organisations (opens in a new tab) to patch actively exploited Citrix NetScaler vulnerabilities. For developers, GitHub has highlighted key changes in Git 2.56 (opens in a new tab), while WooCommerce has moved its experimental Dual API (opens in a new tab) into a standalone plugin.