'Five Eyes' intelligence alliance warns that new AI models pose urgent cyber risk

According to a report from the alliance, these models are expected to exceed current industry expectations, fundamentally altering both offensive and defensive cyber capabilities. The report specifically names Anthropic's Mythos and OpenAI's GPT-5.5-Cyber as examples of models that could be used for complex and potentially devastating hacks.

The alliance's assessment highlights that the rapid advancement of AI models presents an urgent cybersecurity challenge. The report does not call for a halt in development but emphasises the need for coordinated risk management and vigilance among member nations. It underscores the dual-use nature of such technologies, where the same capabilities that drive innovation can also be weaponised.

The warning comes amid growing global concern over AI safety and regulation. The Five Eyes' stance may influence policy discussions in member countries, potentially leading to increased scrutiny of model deployments and cross-border data flows. However, the report does not propose specific new regulations, focusing instead on shared awareness and information sharing.