100+ Firms Warn AI Cyberattacks Set to Surge

More than 100 companies — spanning cloud providers, banks, insurers, security vendors and AI developers — signed an open letter published on August 27, 2026, warning that organizations have only a limited window to shore up their defenses before AI-enabled cyberattacks scale up sharply in the coming months.

Signatories include OpenAI, Anthropic, Microsoft, Google, Amazon Web Services, Oracle, IBM, Cisco, CrowdStrike, Palo Alto Networks, Cloudflare, Okta, Capital One, Mastercard, Visa and Citigroup. "We have a limited window to strengthen cyber defenses," the letter states, calling for coordinated action across industry and government.

The group singled out hospitals, water treatment and wastewater utilities, critical manufacturing and core internet infrastructure as the most exposed targets, arguing that AI models are drastically cutting the time and expertise attackers need to prepare sophisticated intrusions. As evidence, the letter points to an August 18 advisory from the NSA, CISA and the FBI describing AI-generated exploitation scripts, disguised as legitimate software tools, that were used to target Siemens industrial control systems. Separately, CrowdStrike has reported that attackers seized on 88% of newly disclosed proof-of-concept exploits within 48 hours during the first half of 2026.

The letter asks organizations to prioritize eliminating high-risk vulnerabilities and to raise coding standards for AI-generated software, urges vendors to test their products against current AI models and share threat intelligence, and calls on governments to coordinate internationally and fund protection for essential services that often lack dedicated cybersecurity budgets. It also asks AI developers to control how their models can be used for attacks and to provide support during active incidents.

The initiative has drawn skepticism. Reporting on the letter notes it includes no deadlines, spending commitments or measurable targets, and does not call for slowing the pace at which frontier AI models grow more capable — even as some of the same signatories continue releasing increasingly powerful systems. One security researcher likened the dynamic to "an arsonist selling fire extinguishers."

Disclosure: NewUJ's editorial process uses Anthropic's Claude models.