OpenAI Launches Astra With New Cyber and Coding Capabilities

OpenAI has released Astra, describing it as its most powerful and capable model to date and a major advance in computer and browser use. Initial access is going to customers in the Daybreak cybersecurity program, followed by paid ChatGPT plans and the API. OpenAI says Astra improves software engineering, terminal work, bug finding, and cybersecurity tasks while adding safeguards for its stronger capabilities. The release is also controversial because Astra uses a reasoning technique described as opaque recurrence, which critics say can make internal reasoning harder to monitor. Chief scientist Jakub Pachocki acknowledged that oversight becomes more difficult as models grow more capable. President Greg Brockman said he personally believes the industry may have reached AGI, while also noting that AGI no longer has its former contractual role in OpenAI’s partnership with Microsoft.

TechCrunch AI, Lucas Ropek – OpenAI launches Astra, its powerful (and controversial) new model – 3rd September 2026