Five Eyes alliance warns AI could outpace cybersecurity within months

The Five Eyes intelligence alliance cautioned that the fastest-growing artificial-intelligence models could defeat existing cybersecurity defenses in a few months, rather than years. The warning follows a claim by AI startup Anthropic that its Mythos models can locate software vulnerabilities more effectively than prior tools.

In a joint statement released Monday, the security agencies of Britain, the United States, Australia, Canada and New Zealand urged governments and businesses to act quickly as AI capabilities evolve. > "The rapid pace of frontier AI development means cyber risk assumptions can become outdated in months, not years," the statement said.

The advisory noted that AI lowers barriers for malicious actors and accelerates the speed and complexity of attacks, increasing the likelihood of breaches. It added that preparedness can help contain incidents and prevent them from escalating into major operational and financial crises.

To strengthen defenses, the alliance recommended that organizations incorporate AI tools into security operations, modernize legacy systems and restrict access to critical infrastructure. Anthropic recently halted access to its Mythos 5 model and a limited version called Fable 5 after receiving a U.S. national-security directive that barred foreign nationals from using the models. The move comes as the White House has been advocating for reduced AI regulation at the state level.