Cybersecurity Policy Report, Cyber Threat From AI Outstrips Expectations, U.K. Government Warns, (Apr 16, 2026)
A “new generation” of AI (artificial intelligence) models poses a cyber threat that every organization should take seriously, the government of the United Kingdom warned yesterday.
“For years, the most serious cyber attacks have relied on a small number of highly skilled criminals. That is now shifting,” according to an open letter to U.K. businesses from Liz Kendall, head of the U.K.’s Department of Science, Innovation, and Technology (DSIT), and Security Minister Dan Jarvis.
“A new generation of AI models are becoming capable of doing work that previously required rare expertise: finding weaknesses in software, writing the code to exploit them, and doing so at a speed and scale that would have been impossible even a year ago,” Ms. Kendall and Mr. Jarvis noted.
Anthropic’s recently unveiled Claude Mythos2 model was tested by DSIT’s AI Security Institute (AISI) and found to be “substantially more capable at cyber offence than any model we have previously assessed,” they said.
“Recent tests of advanced AI models, including the AISI’s evaluation of Anthropic’s Mythos, indicate that AI cyber capabilities are accelerating even faster than had been previously envisaged,” the officials reported.
“The AISI assess that frontier model capabilities are doubling every 4 months, compared to every 8 months previously. This finding is significant both for what it means today, but also because it highlights the speed at which AI capabilities are increasing and the threats they potentially pose,” they said.
They reiterated that private-sector organizations should avail themselves of resources provided by the U.K. government, including cyber defense guidance and cyber threat alerts.
“We are entering a period in which the pace of technological change may test every institution in the country. The businesses that act now—that treat cyber security as an essential part of running a modern company, not an optional extra—will be the ones best placed to thrive through it and seize its advantages,” Ms. Kendall and Mr. Jarvis stressed.
The U.K. government is not alone in being alarmed about the capabilities of Anthropic’s Mythos2. Anthropic has decided that the model is too advanced to be released publicly and instead should be provided to a limited number of technology firms and organizations and used to find and patch software vulnerabilities (CPR, April 8).
News: InternationalLegislation DataSecurity AINews