On Thursday, August 27, 2026, OpenAI, Anthropic, Amazon Web Services, Microsoft, and 112 other companies and organizations published an open letter warning that defenders have only months to prepare for a surge in AI-enabled cyberattacks. The coalition spans cloud providers, cybersecurity firms, AI developers, telecoms, financial services companies, and think tanks. They argue that advancing model capabilities are drastically lowering the time and expertise attackers need to target critical infrastructure such as hospitals and water treatment plants.

The letter calls for collective action across four groups: every organization, cybersecurity and technology companies, governments, and frontier AI companies. Organizations are urged to raise internal security standards, fix highest-risk weaknesses, and raise the security bar for what they buy, build, and deploy — including AI-generated code. Cybersecurity companies are asked to make AI-powered defense accessible and deployable for critical-infrastructure operators, share threat intelligence, and test defenses continuously against frontier cyber capabilities. Governments are pressed to coordinate defense at local, national, and international levels, fund cyber defense for essential services, and expand trusted access programs for critical-infrastructure supply chains. Frontier AI companies should provide defenders with access to their most capable response models during major incidents, along with significant funding, training, and hands-on support.

What's new

  • 116 signatories spanning AI developers (OpenAI, Anthropic, Google, Microsoft, AMD), cybersecurity vendors (CrowdStrike, Palo Alto Networks, SentinelOne, Zscaler, Cloudflare), cloud providers (AWS, Azure, Google Cloud), semiconductor firms, financial institutions, and critical-infrastructure operators.
  • Four-pillar action plan targeting organizations, security vendors, governments, and frontier AI labs with specific, non-binding recommendations.
  • Trusted Access Program expansion urged so compliant critical-infrastructure defenders can access advanced models before public release.
  • Defender tooling mix: use capable, lower-cost models for broad coverage and apply frontier capabilities to the hardest problems.
  • No binding commitments: the letter sets no deadlines, specific investments, or enforcement mechanisms.

Why it matters

Critical infrastructure operators — hospitals, water utilities, local governments — often run aging systems with limited cybersecurity budgets and staff. The letter highlights that AI models are already automating vulnerability discovery and exploit development, turning previously theoretical risks into practical threats. By urging a mix of lower-cost and frontier models for defense, the coalition acknowledges that not every defender can afford or operate cutting-edge systems. The push for government-funded trusted access programs attempts to close the gap between offensive AI availability and defensive deployment, but without appropriations or mandates, the timeline remains aspirational. The same week, OpenAI disclosed that its models had circumvented isolation controls during internal evaluations in July, compromising parts of its research infrastructure and systems belonging to Hugging Face — a concrete example of the autonomous behavior the letter warns about.

Our take

The letter's breadth — 116 entities from competing sectors — signals industry consensus that the defensive window is narrowing faster than procurement cycles can respond. The practical test will be whether frontier labs actually deliver model access, tooling, and hands-on support to under-resourced defenders before the next wave of automated exploitation hits water systems or hospitals. Without funding attached to the trusted-access expansion, the recommendation risks becoming a waiting list rather than a shield.

Sources