tech
April 15, 2026
OpenAI expands access to cyber AI as hacking risks grow
Facing similar risks as Anthropic, OpenAI is trying a different strategy.
TL;DR
- OpenAI is expanding access to AI models with advanced cyber capabilities by implementing user verification and monitoring systems.
- A new model variant, GPT-5.4-Cyber, has been released to assist with defensive cybersecurity tasks and will be more permissive for vetted users.
- The company's strategy shifts focus from restricting model capabilities to verifying user access to sensitive functions.
- Access will be gradually scaled through OpenAI's Trusted Access for Cyber program, with higher verification tiers unlocking more powerful capabilities.
- GPT-5.4-Cyber is designed to reduce friction for legitimate security work, addressing issues where earlier models refused dual-use cyber queries.
- OpenAI is not currently offering GPT-5.4-Cyber access to U.S. government agencies but is in discussions.
- The speed at which AI models identify security flaws is a growing concern for government officials and business leaders.
Continue reading the original article