OpenAI Hosts Summit with 300 Security Leaders to Expand AI Access for Defenders
OpenAI President Greg Brockman is convening a cybersecurity summit with around 300 security leaders and Chief Information Security Officers from Fortune 100 companies and government agencies. At the summit, OpenAI plans to expand access to its AI tools specifically for defenders protecting public sector institutions and critical infrastructure. As frontier AI models rapidly gain advanced offensive cyber capabilities, giving infrastructure defenders equal access to AI tools is critical to maintaining network defenses. This initiative reflects a growing urgency to equip public utilities and essential services against impending AI-powered cyberattacks. The announcement comes alongside internal concerns over OpenAI's upcoming Astra model, which reached the 'Critical' capability threshold under OpenAI's Preparedness Framework. Additionally, security professionals have raised alarms over new AI training methods that may obscure the reasoning processes of models, making auditing more difficult.
## BACKGROUND
OpenAI's Preparedness Framework tracks and mitigates potential catastrophic risks from frontier AI models across categories like cybersecurity, threat creation, and autonomy. Reaching a 'Critical' threshold in cybersecurity indicates a model possesses capabilities that could facilitate serious cyberattacks, triggering stricter access controls and safeguards before release.