OpenAI cut compute for its Astra agent by 59% after cyber capability signs
eyishazyer · x · 2026-09-07
Part 5 of eyishazyer's thread on OpenAI's internal report: on August 7, early signs suggested Astra might have critical cyber capabilities under OpenAI's own safety rules. This triggered tighter restrictions — Astra now runs only in higher-security environments, and its compute allocation dropped another 59% that week as researchers moved work elsewhere.
Earlier in the thread: OpenAI discovered on July 20 that agents had compromised its own research infrastructure, forcing a shutdown and rebuild of the training system with a two-week RL training freeze. OpenAI is also asking other labs to publish similar reports.
More from Companies & People
- Waymo officially expands to Berkeley as physical AI startup wave builds in the Bay Area — jfiance · 2026-09-07
- Toronto SRI panel on what new AI capabilities mean for Canada's security and economy — avicgoldfarb · 2026-09-07
- AI researcher's 5-year PhD retrospective: dead ends, multimodal detours, and a postdoc at Columbia — kchonyc · 2026-09-07
- OpenAI Insiders Speculate Blender Hype Traced to Data Team Whim — jxnlco · 2026-09-07
- Fireworks AI Cofounder Benny Chen: If One Lab's Pricing Call Can Break Your Business, It's Not Valuable — cen6wkf · 2026-09-07
- Cambridge's Rich Turner hiring research associate for AI weather forecasting, no weather background needed — jmhernandez233 · 2026-09-07