Ofcom writes to broadband providers following Anthropic Mythos AI concerns

Natalie Black, Group Director for Infrastructure and Connectivity at Ofcom has written to broadband providers to highlight concerns about the “rapid increase in AI ability”, specifically relating to Anthropic’s Claude Mythos Preview AI model which has been widely reported on in the past few weeks as representing a fundemental shift to security capabilities. The model is not currently available for wider use.

The AI Security Institute said the model represented “a step up over previous frontier models in a landscape where cyber performance was already rapidly improving”.

“I am writing to highlight the urgent challenge posed by the rapid increase in AI capability. You will be aware that Anthropic announced Claude Mythos Preview, a new frontier model with significantly advanced cyber capabilities.

DSIT’s AI Security Institute published its evaluation of Mythos and noted that this model is incrementally more capable in terms of exploitation of vulnerabilities than any it has previously assessed. This marks a clear escalation in capability and forms part of a broader and rapidly accelerating trend. Frontier models will continue to advance quickly, increasing both the scale and speed of cyber threats. Continued and proactive investment in cyber defence is critical.

Last week, the Secretary of State for Science, Innovation and Technology Liz Kendall MP, and Security Minister Dan Jarvis MP set out their concerns and expectations in their open letter to business leaders. They make clear that the growing challenge for cyber defence is not isolated to a single company and we should be prepared for frontier AI capability to rapidly increase over the next year.

Against this backdrop, you should assess security risks arising from frontier AI models and take appropriate mitigating action, in line with your security duties and existing regulatory guidance. As a Communications Provider we expect you to implement appropriate measures, in line with those in DSIT’s Telecommunications Code of Practice. In addition, we expect you to follow any relevant advice from NCSC on this issue, including their recent published note on the cybersecurity challenges of frontier AI. This provides helpful information to guide companies’ response to this developing risk.

Ofcom is closely monitoring these technical developments, and we will take a proactive approach to working with you to address the challenges and opportunities. We will be in touch to take stock of progress so far and where additional action may be required.”

Natalie Black, Ofcom (our emphasis)