Global Security Agencies Warn of Rapid AI Threats

Cyber intelligence agencies from the Five Eyes alliance—a group including Australia, the US, the UK, New Zealand, and Canada—have issued a rare and urgent warning. They claim that new, powerful artificial intelligence models capable of causing significant harm to governments and businesses could be ready in just months. The agencies are calling on world leaders to take immediate action to prepare for these changes.

According to The Guardian, the Five Eyes statement highlights that while AI can help improve cyber defense, it also makes cyber threats much faster and more complex. The agencies noted that the speed of these developments is not measured in years, but in months. They believe that these "frontier" AI models—the most advanced systems currently being built—will change how cyberattacks work, making them harder to stop. The statement emphasized that cyber risk is no longer just a technical problem for IT departments; it is now a core business risk that requires attention from top leaders.

This warning follows a recent decision by the US government to block "foreign nationals" from using two advanced AI models created by the company Anthropic, known as Mythos and Fable. Mythos is a powerful tool capable of finding weaknesses in computer systems. Because of these risks, it is only available to vetted, or carefully checked, organizations. The US government made this decision after receiving advice from national security officials.

There is a growing debate about how these AI companies communicate the dangers of their products. According to Ars Technica, some critics believe that Anthropic has been too focused on highlighting risks. Research shows that Anthropic uses words related to danger and regulation much more often than its rival, OpenAI. Some experts, such as Meta’s former chief AI scientist Yann LeCun, have criticized this approach. LeCun suggested that the company’s repeated warnings about AI risks have backfired, leading directly to the government’s decision to restrict access to their technology.

Olivia Shen, an expert in national security and AI at the University of Sydney’s United States Studies Centre, told The Guardian that the world is currently focused on Anthropic, but there are likely many other powerful models being developed elsewhere. Shen noted that we should expect new, advanced models to appear soon, potentially from other countries or different companies that are not currently under the same level of public scrutiny.

This situation is emerging as a major test for how the US government intends to oversee the AI industry. While the US is moving toward stricter controls, other nations are taking different paths. For example, the Australian government has signed a memorandum of understanding—a formal agreement that is not legally binding—with Anthropic. This agreement encourages the company to share information about its progress and work with the government to keep AI safe, while still allowing for a "light-touch" approach to regulation. This strategy aims to help the country gain economic benefits from the technology without slowing down innovation too much.

As the technology evolves, the challenge for leaders remains the same: how to capture the benefits of AI while protecting society from the risks. The Five Eyes agencies have made it clear that a "whole-of-society" response is needed. This means that governments, businesses, and individuals must all work together to build what they call "cyber resilience"—the ability to prepare for, respond to, and recover from cyber threats. As the speed of AI development continues to increase, the pressure on leaders to find the right balance between safety and progress will only grow stronger.