‘Act now’: Five Eyes warns that AI models specialized for cyber attacks are only months away

Five Eyes alliance warned cross-border GenAI models will enable advanced cyber attacks against companies and governments within months
Statement emphasized that cyber risk is now a management and business continuity issue that requires a community-wide response
Comes amid concerns over Anthropic’s Mythos Preview and other models already showing offensive potential despite auto guards

In just a few months, advanced Generative Artificial Intelligence (GenAI) models will be able to launch cyber attacks on large companies and government organizations, warns Five Eyes.

The Five Eyes is an intelligence-sharing alliance between the United States, Great Britain, Canada, Australia and New Zealand. It was formed after World War II and allows the five countries to cooperate closely on intelligence and national security issues.

Earlier this week, Five Eyes issued a new warning that said AI will help improve cyber defenses over time, but will also accelerate the speed, scale and sophistication of threats: “Frontier AI models are expected to exceed current industry expectations, fundamentally transforming both offensive and defensive cyber capabilities. The timeline is not months,” the warning says. “In this environment, cyber resilience is integral to promoting business continuity, market confidence and long-term value.”

All hands on deck

Five Eyes now says the industry needs all hands on deck to address what is increasingly becoming a pressing issue:

“An organization-wide and community-wide response is required,” it said. “Cyber risk can no longer be treated as a purely technical issue. This is a core business risk and management responsibility.”

In early April, news broke that Anthropic’s latest AI model, Mythos Preview, was so good at exploiting software vulnerabilities that the company couldn’t release it to the public. Instead, it only shared it with a handful of US companies to give them an edge against threat actors.

While skeptics said it was nothing more than a publicity stunt, similar to what OpenAI achieved with ChatGPT 2.0, companies using it (for example Mozilla) confirmed that it was indeed strong enough to be contained.

Even models available today, despite all the safeguards, are regularly exploited by bad actors in various cyber attack scenarios.

Via The Guardian