Intelligence agencies warn AI models could launch crippling cyberattacks in months

AI firms have previously cautioned about the technology's potential to empower malicious actors to hack faster, cheaper, and at a broader scale. These concerns have intensified this year with the development of new AI cybersecurity models, such as Anthropic's Mythos model, which the company withheld from full public release due to its perceived danger. The agencies noted that AI lowers barriers for malicious actors and increases the speed and complexity of attacks, shrinking the window between vulnerability discovery and exploitation.

The warning targets small and medium businesses, large organizations, critical infrastructure, and governments, all of which are considered at risk. The intelligence agencies recommended that leaders accelerate patching processes, address unsupported legacy systems, and prepare for incidents proactively. They also advised implementing secure-by-design and secure-by-default practices.

In the U.S., the White House has been deliberating on how to manage the cybersecurity risks associated with newer AI models, seeking to balance safety concerns with a preference for light-touch regulation. President Trump recently signed an executive order establishing a voluntary testing process where AI labs can submit their models to the government for risk assessment up to 30 days before public release. However, some observers predicted these voluntary assurances would be insufficient. Weeks later, the administration directed Anthropic to withdraw its Fable and Mythos models, a move that generated significant backlash from the AI policy community.