This statement did not emerge from a vacuum. It was the culmination of a rapid cascade of events over the preceding two months — a chain that included a government shutdown of Anthropic's most powerful models, a startlingly simple exploit discovered by a cybersecurity researcher, the launch of a major rival cybersecurity platform by OpenAI, and an earlier foundational advisory from the Five Eyes themselves on autonomous AI agents.

The Immediate Trigger: Anthropic's Fable 5 Takeover

The most direct catalyst for the June 22 warning was the dramatic rise and fall of Anthropic's Claude Fable 5.

On June 9–10, 2026, Anthropic released two models: Claude Fable 5, a powerful general-purpose AI model available to the public, and Claude Mythos 5, the same underlying model with cyber safeguards intentionally removed, restricted to a vetted group of cyber defenders and critical infrastructure operators . Anthropic described Mythos 5 as "the strongest cybersecurity model in the world" .

The lifespan of these models in the open was measured in days. On June 12 at 5:21 p.m. ET, the US Commerce Department issued an emergency export control directive. The order required Anthropic to immediately disable both models for any foreign national, including foreign employees of Anthropic itself . Anthropic had no practical way to filter access by nationality, so it suspended global access for all users that same day .

The 'Fix This Code' Vulnerability

The trigger for the government action: cybersecurity researcher Katie Moussouris demonstrated that the models could be prompted with the simple three-word instruction "Fix this code" to autonomously reverse-engineer and weaponize software vulnerabilities, raising urgent national security alarms . The exploit was discovered after Amazon, an Anthropic investor, conducted its own security review and alerted the White House . Anthropic had asked Moussouris to review a report on the vulnerability produced by Amazon researchers .

Earlier Warning Sign: Mythos Unauthorized Access

The Fable 5 incident was not the first sign of trouble. In April 2026, Anthropic itself investigated allegations that a small group had gained unauthorized access to an earlier Claude Mythos preview model — a system the company described as too potent for public release . This incident foreshadowed the broader concerns about proliferation risks that would later trigger the Fable 5 shutdown .

The Competitive Response: OpenAI's Daybreak Platform

A month before Fable 5's launch, OpenAI had already made its own major move into the AI cybersecurity space.

On May 11, 2026, OpenAI launched Daybreak, a major cybersecurity initiative . The launch occurred on the same day Google's Threat Intelligence Group disclosed the first confirmed case of attackers using AI to build a zero-day exploit — a 2FA bypass that a prominent cybercrime group planned to use in a mass exploitation campaign .

Daybreak bundles OpenAI's GPT-5.5 model tiers — including a permissive "GPT-5.5-Cyber" red-teaming variant — with Codex Security agents and over 20 security partners (including Cloudflare, CrowdStrike, Palo Alto Networks, Cisco, Oracle, and others) to automate vulnerability discovery, patch validation, and remediation at machine speed . It was widely seen as OpenAI's direct competitive response to Anthropic's earlier Project Glasswing / Mythos offerings .

The Policy Foundation: Five Eyes Agentic AI Advisory

The Five Eyes had already laid the policy groundwork for the June 22 warning a month earlier. On May 1, 2026, six cybersecurity agencies across the Five Eyes (CISA, NSA, ASD ACSC, CCCS, NZ NCSC, UK NCSC) published "Careful Adoption of Agentic AI Services" . This was the first coordinated multi-government security guidance specifically targeting autonomous AI agents — AI systems that can plan, make decisions, and take actions over extended periods without human approval at each step .

The advisory identified five risk categories: privilege, design/configuration, behavioral, accountability, and data/operational risks . It warned that autonomous agents that "act, decide, and interact across systems in ways that are difficult to predict, audit, or control" were already operating in critical infrastructure with insufficient safeguards . This May advisory laid the policy groundwork that the June 22 statement escalated into a broader, more urgent call to action.

Summary Timeline

The May agentic AI advisory identified the systemic risks. The Mythos access incident and the "Fix this code" vulnerability in Fable 5 proved that those risks were already materializing in real-world frontier models. OpenAI's Daybreak launch intensified the competitive and security dynamics. The June 22 statement represents the Five Eyes' unified conclusion: the window for defensive action has collapsed from years to months.

This statement did not emerge from a vacuum. It was the culmination of a rapid cascade of events over the preceding two months — a chain that included a government shutdown of Anthropic's most powerful models, a startlingly simple exploit discovered by a cybersecurity researcher, the launch of a major rival cybersecurity platform by OpenAI, and an earlier foundational advisory from the Five Eyes themselves on autonomous AI agents.

The Immediate Trigger: Anthropic's Fable 5 Takeover

The most direct catalyst for the June 22 warning was the dramatic rise and fall of Anthropic's Claude Fable 5.

On June 9–10, 2026, Anthropic released two models: Claude Fable 5, a powerful general-purpose AI model available to the public, and Claude Mythos 5, the same underlying model with cyber safeguards intentionally removed, restricted to a vetted group of cyber defenders and critical infrastructure operators . Anthropic described Mythos 5 as "the strongest cybersecurity model in the world" .

The lifespan of these models in the open was measured in days. On June 12 at 5:21 p.m. ET, the US Commerce Department issued an emergency export control directive. The order required Anthropic to immediately disable both models for any foreign national, including foreign employees of Anthropic itself . Anthropic had no practical way to filter access by nationality, so it suspended global access for all users that same day .

The 'Fix This Code' Vulnerability

The trigger for the government action: cybersecurity researcher Katie Moussouris demonstrated that the models could be prompted with the simple three-word instruction "Fix this code" to autonomously reverse-engineer and weaponize software vulnerabilities, raising urgent national security alarms . The exploit was discovered after Amazon, an Anthropic investor, conducted its own security review and alerted the White House . Anthropic had asked Moussouris to review a report on the vulnerability produced by Amazon researchers .

Earlier Warning Sign: Mythos Unauthorized Access

The Fable 5 incident was not the first sign of trouble. In April 2026, Anthropic itself investigated allegations that a small group had gained unauthorized access to an earlier Claude Mythos preview model — a system the company described as too potent for public release . This incident foreshadowed the broader concerns about proliferation risks that would later trigger the Fable 5 shutdown .

The Competitive Response: OpenAI's Daybreak Platform

A month before Fable 5's launch, OpenAI had already made its own major move into the AI cybersecurity space.

On May 11, 2026, OpenAI launched Daybreak, a major cybersecurity initiative . The launch occurred on the same day Google's Threat Intelligence Group disclosed the first confirmed case of attackers using AI to build a zero-day exploit — a 2FA bypass that a prominent cybercrime group planned to use in a mass exploitation campaign .

Daybreak bundles OpenAI's GPT-5.5 model tiers — including a permissive "GPT-5.5-Cyber" red-teaming variant — with Codex Security agents and over 20 security partners (including Cloudflare, CrowdStrike, Palo Alto Networks, Cisco, Oracle, and others) to automate vulnerability discovery, patch validation, and remediation at machine speed . It was widely seen as OpenAI's direct competitive response to Anthropic's earlier Project Glasswing / Mythos offerings .

The Policy Foundation: Five Eyes Agentic AI Advisory

The Five Eyes had already laid the policy groundwork for the June 22 warning a month earlier. On May 1, 2026, six cybersecurity agencies across the Five Eyes (CISA, NSA, ASD ACSC, CCCS, NZ NCSC, UK NCSC) published "Careful Adoption of Agentic AI Services" . This was the first coordinated multi-government security guidance specifically targeting autonomous AI agents — AI systems that can plan, make decisions, and take actions over extended periods without human approval at each step .

The advisory identified five risk categories: privilege, design/configuration, behavioral, accountability, and data/operational risks . It warned that autonomous agents that "act, decide, and interact across systems in ways that are difficult to predict, audit, or control" were already operating in critical infrastructure with insufficient safeguards . This May advisory laid the policy groundwork that the June 22 statement escalated into a broader, more urgent call to action.

Summary Timeline

The May agentic AI advisory identified the systemic risks. The Mythos access incident and the "Fix this code" vulnerability in Fable 5 proved that those risks were already materializing in real-world frontier models. OpenAI's Daybreak launch intensified the competitive and security dynamics. The June 22 statement represents the Five Eyes' unified conclusion: the window for defensive action has collapsed from years to months.