OpenAI's Next AI Model Could Enable Fully Autonomous Cyberattacks, Experts Warn

OpenAI's current model, GPT-5.6-Sol, has already been classified as a 'high' cybersecurity risk. Its successor, reportedly codenamed Astra, is expected to potentially reach the 'critical' threshold — the highest risk category used to assess how capable an AI system might be of assisting or independently carrying out cyberattacks.

While details remain limited, the concern centres on AI models becoming sophisticated enough to plan and execute attacks with little or no human involvement. This could lower the skill barrier for cybercriminals, making advanced attack techniques more accessible and harder to predict or defend against using traditional security measures.

For small and medium businesses, this development is a signal that the threat landscape is evolving quickly. As AI tools become more powerful, both defenders and attackers will have access to increasingly capable technology, meaning businesses should not assume that current security practices will remain sufficient indefinitely.