OpenAI has released ChatGPT Cyber, a new cybersecurity model aimed at helping defenders find and patch software vulnerabilities. The release follows recent demonstrations where advanced AI models successfully uncovered severe bugs in major open-source systems, including the Linux kernel and the Firefox browser. Max Justicz, cofounder of Charge Robotics, used an earlier OpenAI model to discover a severe vulnerability in Ubuntu servers within just two hours, highlighting how frontier models can assist with complex security audits.
In response to these emerging risks, OpenAI has expanded its Daybreak security initiative to provide advanced tools to researchers and open-source maintainers. The company announced a $1 billion initiative to support cybersecurity professionals protecting critical infrastructure such as electrical grids and water systems. While these programs aim to empower defenders, critics and researchers note that balancing safety guardrails with effective vulnerability research remains an ongoing challenge for the industry.
The rapid advancement of AI-driven hacking tools underscores an urgent race between defenders and malicious actors to secure digital infrastructure.