According to OpenAI, GPT-6 Astra is the first model to reach the Critical level for cybersecurity capabilities under the company’s Preparedness Framework.
With appropriate tools and access, the model can identify previously unknown security vulnerabilities and discover new ways they could potentially be exploited, prompting OpenAI to introduce additional safeguards.
The new AI model is capable of carrying out multi-step computer tasks, including working with software and browsers, creating spreadsheets, building websites, conducting online research and completing various professional workflows.
OpenAI President and co-founder Greg Brockman described the model as a major step forward in AI’s ability to use computers.
The company said GPT-6 Astra has demonstrated significant improvements over its predecessor, GPT-5.6 Sol, in computer use, cybersecurity and other complex tasks.
However, the increased capabilities of the new model have also raised security concerns. OpenAI acknowledged that monitoring increasingly autonomous and powerful AI systems and keeping them within human-defined limits could become a major challenge in the future.
According to the company, GPT-6 Astra is initially being rolled out to a limited number of organisations. It is expected to become more widely available in the coming days to ChatGPT Plus, Pro, Business and Enterprise users, as well as developers.