AI-Powered Cyberattacks on Advanced Systems May Be Months Away, Intel Agencies Warn

Advanced artificial intelligence (AI) models may be able to breach and overwhelm the cybersecurity defenses of governments and businesses worldwide in a matter of months, a partnership of intelligence agencies warned on June 22.

The Five Eyes Intelligence Oversight and Review Council (FIORC), an alliance of intelligence and security agencies from the United States, the United Kingdom, Canada, Australia, and New Zealand, issued a joint statement that raises alarms over the threats of frontier AI models.

“Frontier AI models are anticipated to exceed current industry expectations, fundamentally transforming both offensive and defensive cyber capabilities. The timeline is not years, it is months,” the statement said.

FIORC warned world leaders and businesses to assess the risks of frontier AI to respond quickly and prevent “malicious actors” from exploiting software breaches that could impact national defense.

“Boards and executives should ensure cyber resilience is in place and works under pressure. It is not enough to have controls. Leaders must be confident those controls will perform during a real incident,” the statement said.

“This requires reassessing long-standing trade-offs and using AI deliberately to strengthen defense—not just improve efficiency.”

Monday’s statement from FIORC did not reference any particular AI firms, but recent industry discussion has focused on Anthropic’s Mythos 5 and Fable 5 frontier models.

The unusual joint statement comes a little more than a week after the Trump administration ordered Anthropic to ban foreign nationals worldwide from accessing its Claude Fable 5 and Mythos 5 frontier models. Due to the difficulty of determining which users would fit this broad category, the AI firm pulled access to all users shortly after.

However, Anthropic said the federal government believed that someone had found a way to “jailbreak” Fable 5, which would essentially mean removing the security safeguards that prevent the model from being as powerful as Mythos 5.

When Anthropic initially announced a limited release of Mythos to select cybersecurity testers worldwide in April, the company warned that bad actors could use the model to find exploits in operating systems, banks, and other critical infrastructure.

Prior to Anthropic pulling access to both models for all users on June 12, Mythos 5 was still in a limited rollout to select companies, firms, and cybersecurity testers.

Then, days later, President Donald Trump told Axios that he no longer viewed Anthropic as a national security threat due to how quickly the company responded to his requests.

Trump also said that someone working for one of Anthropic’s competitors in the AI industry appeared to be concerned over the new models and “turned Anthropic in.”

Before initially releasing those models earlier this month, Anthropic called on AI industry leaders worldwide to slow the development pace of the technology out of fears that it could soon start advancing itself beyond humanity’s control.

This is due to firms using AI itself to design, develop, and enhance the models, which could soon lead to an “AI system capable of fully autonomously designing and developing its own successor,” Anthropic warned.