The Five Eyes raise the alarm: AI cyberattacks accelerate to months, not years
Cybersecurity agencies of the Five Eyes countries — Australia, Canada, New Zealand, the United Kingdom, and the United States — have issued a joint statement warning that advanced artificial intelligence models are fundamentally reshaping the cyber threat landscape. The pace of these changes is measured not in years, but in months. This is not merely a forecast, but a statement of fact that demands an immediate reassessment of risk management approaches in both business and government structures.
What exactly has changed?
According to the document, AI is already actively accelerating, scaling, and complicating cyberattacks. The technology lowers the barrier to entry for malicious actors: conducting sophisticated attacks no longer requires years of experience. At the same time, the time gap between vulnerability discovery and exploitation is shrinking. However, AI is a double-edged sword. The same models can help defenders find weaknesses faster, improve code quality, and automate incident response.
The key conclusion of the Five Eyes: cyber risks are no longer purely a technical problem. This is a matter of operational resilience, market trust, and direct management responsibility. Boards of directors are recommended not just to have defense systems in place, but to be confident they can withstand a real attack.
From theory to practice: AISI tests and new models
These warnings are not abstract. Assessments by the UK AI Security Institute (AISI) reveal alarming dynamics. While in March 2026 no public model could fully complete a 32-step simulation of an attack on a corporate network, by April the Claude Mythos Preview model (from Anthropic) succeeded in 3 out of 10 attempts. By May, the same model improved its result to 6 out of 10 attempts and, for the first time, completed a second, more complex scenario.
The cost of one such attack in the simulation is estimated at just £65. This demonstrates how accessible a powerful cyber tool has become. It is important to note that the tests were conducted under simplified conditions — without active defenders — which certainly overestimates AI capabilities, but the direction and speed of progress are evident.
Industry response: GPT-5.5-Cyber and Daybreak
Alongside the warnings, major market players are stepping up. OpenAI has released the full version of the GPT-5.5-Cyber model, designed for verified security professionals. The model achieved a record efficiency of 85.6% on the CyberGym platform. Additionally, the Daybreak initiative has been launched — a set of tools for automated vulnerability discovery and remediation, along with the Patch the Planet project in collaboration with HackerOne to assist open-source projects.
Analyst's opinion
We are witnessing a classic digital arms race, but at an unprecedented speed. AI agents capable of autonomously planning and executing multi-step attacks are becoming a reality. For the crypto industry, where smart contract security and transaction privacy are foundational, this means that old audit and protection methods may become obsolete within months. Investors and developers should pay close attention to solutions that integrate AI for proactive defense, rather than just post-mortem analysis.