6 September 2026
OpenAI developer says Astra agent boosted internal productivity by six months
First reported
TechRadar ran this on , a day before the next source picked it up.
- Astra is OpenAI's new AI agent model, designed to autonomously perform tasks on computers. An OpenAI developer claimed using it internally accelerated project timelines significantly.
- A survey of 25 researchers across OpenAI, Anthropic, Google DeepMind, and Meta found 20 ranked AI systems automating research itself as a major risk to monitor.
- Anthropic reports Claude now writes over 80 percent of its own production code, suggesting similar productivity gains from autonomous AI at that company.
- OpenAI says Astra includes monitoring to catch the model misbehaving mid-task, but security researchers warn its reasoning is increasingly hidden from human inspection, making audits harder.
- Astra will run as an agent on employee laptops and browsers with real credentials, raising questions about who is liable if an autonomous AI system causes financial loss or leaks data.
Reported by The Decoder, TechRadar