Five Eyes warns AI-driven cyber threats could overwhelm defenses within months

Intelligence agencies of the Five Eyes alliance – the United States, United Kingdom, Canada, Australia and New Zealand – issued a joint statement saying that advanced frontier AI models are expected to enable large‑scale cyber attacks in months, not years. They warned that generative AI lowers barriers for malicious actors, accelerates vulnerability discovery and shortens the window between exploit and impact, threatening critical infrastructure and corporate systems. The agencies urged organisations to treat cyber risk as a core business issue, accelerate patching, enforce zero‑trust access controls, limit system exposure and develop AI‑against‑AI defensive tools.

The statement cited recent U.S. export controls that barred foreign nationals from Anthropic’s Mythos 5 and Fable 5 models, underscoring growing geopolitics around AI access. Chinese firms are responding with cheap, open‑source models such as GLM‑5.2, DeepSeek and others that are rapidly gaining users in the U.S. market, prompting concerns about “distillation” of data from Western models and a potential shift in AI market power. Chinese AI startup Zhipu AI’s founder claimed China could develop a Mythos‑class model before 2027, intensifying the strategic AI race.

These developments highlight a tightening U.S.‑centric AI monopoly debate, with countries like India watching how restrictions could affect their access to leading models. The convergence of AI‑enhanced cyber threats and rapid Chinese AI advancement is prompting worldwide calls for coordinated security and policy responses.