Five Eyes Warn AI Models Can Autonomously Hack Networks as Cyber Risks Shift in Months
Updated
Updated · The Guardian · Jun 29
Five Eyes Warn AI Models Can Autonomously Hack Networks as Cyber Risks Shift in Months
3 articles · Updated · The Guardian · Jun 29
Summary
Five Eyes security agencies said AI systems can now hack networks, steal data, deploy ransomware and damage systems with minimal prompting, sharply widening the pool of potential attackers.
Months—not years—now separate outdated and current cyber-risk assumptions, the agencies said, arguing that rapid frontier-model advances are accelerating threats faster than standard defenses can adapt.
Open-source and locally run models could blunt guardrails imposed by major AI companies, making long-term restrictions on harmful cyber use difficult to enforce.
AI-driven defense is the main remedy the statement highlights: using the technology to find vulnerabilities earlier, improve software quality, detect unusual behavior and speed incident response.
The warning largely reinforces long-standing security advice, but frames it as newly urgent because AI is shrinking the gap between technical skill and the ability to launch damaging attacks.
When AI can hack complex systems with a simple prompt, what is the new role for human cybersecurity experts?
Tech giants are building AI to find flaws. Can we trust them to control these powerful digital skeleton keys?
AI now creates super-hackers and super-defenders. Who is actually winning this escalating cyber arms race?
Autonomous AI Hacking Becomes Reality: 2026 Five Eyes Warning and the Race for Cybersecurity Resilience
Overview
In late June 2026, the Five Eyes intelligence alliance issued a rare joint warning, confirming that autonomous AI hacking had become a real and urgent threat. This shift was driven by the emergence of powerful 'frontier' AI models like Anthropic’s Fable and Mythos and OpenAI’s Daybreak and GPT-5.5, which can now autonomously discover and patch code vulnerabilities. These advanced systems, whose performance has far surpassed previous trends, are being leveraged by adversaries to move faster and more effectively in cyber operations, fundamentally changing the speed, scale, and nature of global cybersecurity risks.