AI news story
OpenAI opens GPT-5.5-Cyber to vetted security researchers
OpenAI is releasing GPT-5.5-Cyber, a model variant that rejects far fewer security requests and even actively executes exploits against test servers. Access is limited to verified defenders of critical infrastructure, including partners like Cisco, C
Editor's take
OpenAI is providing select cybersecurity professionals with access to GPT-5.5-Cyber, a specialized version of its large language model capable of identifying and executing security vulnerabilities within controlled environments.
This move signifies a critical step in the dual-use AI arms race, enabling defenders to proactively understand and counter AI-powered attack vectors. By allowing vetted partners like Cisco to test the model on their infrastructure, OpenAI is fostering a more robust defense against sophisticated threats, particularly those targeting critical national infrastructure.
The success of this program hinges on OpenAI's ability to maintain strict access controls and monitor model usage. Future developments to observe include the emergence of novel exploit techniques discovered by researchers and the potential for similar specialized models to be developed by competing AI labs for both offensive and defensive purposes.
Signal score: 3
This event was corroborated by 23 independent sources. The signal score weighs cross-source corroboration, recency, source weight and topic salience. How we rank stories.
Original reporting
This story summarises reporting published by The Decoder. Read the original article at The Decoder.