Five Eyes warns AI on pace to bypass cybersecurity within months

What happened

The Five Eyes alliance (US, UK, Canada, Australia, New Zealand) issued a joint statement on June 22 warning that frontier AI is progressing fast enough that foundational cybersecurity assumptions could become outdated within months.

Context and impact

The warning follows Anthropic's April disclosure that Mythos had unprecedented vulnerability-finding abilities, and lands a day before OpenAI expanded Daybreak with GPT-5.5-Cyber. The alliance expects attacks to escalate in speed, scale and sophistication while the barrier of entry drops.

Details

• Joint statement issued Monday June 22
• Quote: 'cyber risk assumptions can become outdated in months, not years'
• Background: Anthropic Mythos broke into almost all NSA classified systems in hours (April 2026)
• OpenAI GPT-5.5-Cyber scored a record 85.6% on CyberGym