AI attacks capable of breaching government security could emerge within months, Five Eyes warn

A stark warning has emerged from the heads of intelligence agencies in five Western nations: AI powerful enough to dismantle existing government and corporate security frameworks could appear within months. The "Five Eyes" intelligence-sharing alliance, comprising the United States, United Kingdom, Canada, Australia, and New Zealand, issued a joint statement Tuesday. They cautioned that leading AI models are evolving at a pace that will reshape cyber capabilities far quicker than industry projections. The agencies highlighted that malicious actors will soon find it easier to launch attacks, which will also become faster and more complex. This rapid advancement narrows the window for identifying and exploiting vulnerabilities, necessitating urgent, coordinated responses. The statement follows a U.S. directive last week to block foreign access to Anthropic's advanced AI model, citing security concerns. Financial Times interpreted the warning as an unusual acknowledgment that the West's current advantage in commercial AI development and military/intelligence applications over Russia and China may not last. The cybersecurity chiefs effectively urged Western firms to adopt AI models to bolster defenses, potentially fueling an arms race.

“

Frontier artificial intelligence models will change cyber capabilities at a pace that surpasses current industry expectations, and the timeline is measured in months, not years.