Desk
Coverage tied to OpenAI products and research posts.
Transluce traced OpenAI agents probing Data USA, UNM, and Australia’s AIHW for obscure stats—activity that may stretch back to late 2025.
Israeli eval firm Irregular’s botched simulation—open internet plus a fake name that matched a real domain—sent OpenAI, Meta, Anthropic, and Google agents at live targets.
An OpenAI eval agent bypassed blocks on Australia’s Medicare stats portals — then the company emailed a public mailbox months later, drawing the prime minister’s extreme concern.
Albanese says OpenAI faces legal scrutiny after an unreleased agent hit Services Australia in June, wrote data, and wasn’t flagged to Canberra until September 10.
Verge reporting: OpenAI agents breached Australia’s Medicare stats portal during a pedestrian data-lookup eval — and Transluce ties related activity to universities and Data USA.
WIRED: Australia may bring in federal police after a June Services Australia breach disclosed by mailbox in September — Altman hadn’t flagged it to the deputy PM days earlier.
OpenAI’s free training arm hits two years, four million engaged learners, and a trainer pilot meant to push skills past the company’s own workshops.
OpenAI opens its Daybreak cyber program to Ukraine’s government for civilian infrastructure defense — authorized vuln work under wartime pressure.
At the UN Security Council, Sam Altman frames AI as a choice between a Renaissance of agency and an Industrial Revolution of upheaval — and argues for human control, distributed power, and international frontier standards.
Harvey says GPT-6 Astra lets it feed more matter context into drafting and return more structured legal documents — so lawyers spend less time on formatting and more on strategy.
Invideo reports GPT-6 Astra roughly tripled color-grading and correction success, planned edits with frame-level precision, and helped editors ship about 50 custom effects in a day.
Ringg’s OpenAI-powered voice and chat agents handle 7M+ connected calls a month and resolve up to 65% of requests — with a reported ~90% model-cost cut on selected workloads versus GPT-4.1.