1 of 2
Story summary
- OpenAI intensified cybersecurity after tests show its upcoming AI models have achieved 'high' cyber-task capabilities.
- Capture-the-flag performance rose from 27% to 76% in three months.
- OpenAI is implementing a defense-in-depth strategy with access controls, infrastructure security, monitoring systems, establishing the Frontier Risk Council, and collaborating with other labs on shared threat models.
- OpenAI will launch a trusted access program for qualified users and deploy Aardvark tool to identify code vulnerabilities.
1 / 2
