OpenAI and more than 100 signatories published an August 27 open letter warning that AI-enabled cyberattacks are set to become more widespread and sophisticated. The list includes Anthropic, Google, Microsoft, Cisco, CrowdStrike, Cloudflare, IBM, Oracle and others across AI, cloud, finance and security.
The letter says attackers will use AI to find weak authentication, misconfigurations, unpatched software and legacy technical debt faster, especially in hospitals, water systems and internet infrastructure. It argues that the same AI systems can also lower the cost of defense, if vendors make capable models available to defenders, share threat intelligence, and support critical services with funding, verified playbooks and traceable model access.
For readers running essential services, the practical shift is that AI defense is now being framed as a procurement and operating assumption, not just a tooling choice. The unresolved question is who gets to define those defensive standards: the companies building the models, the vendors selling the controls, or the public agencies paying to keep critical systems online.