AI Cyberattack Risk Could Hit Hospitals Within Months, 100 Firms Warn

AI cyberattack risk drove more than 100 technology and cybersecurity firms, including Microsoft, Google, OpenAI and Anthropic, to warn Washington that more capable AI systems could fuel attacks on hospitals and power grids within months.

Key Takeaways

• More than 100 technology and cybersecurity firms signed a joint letter warning of AI-fueled attacks on hospitals and power grids
• Microsoft, Google, OpenAI, Anthropic, Amazon Web Services and Adobe are among the letter’s signatories
• The letter warns that agentic AI can scan networks, write exploit code and chain intrusion steps without a human operator
• In 2021, an intruder remotely raised sodium hydroxide levels at a Florida water treatment plant in Oldsmar

What The AI Cyberattack Risk Letter Warns

The joint letter, titled “A Call for Collective Action on Cyber Defense,” ties the threat directly to hospitals, electric utilities and water treatment plants. Amazon Web Services and Adobe also signed the document alongside the four largest AI labs. Local outlets carried the letter’s warning within hours of its Friday release, citing the same list of targets in a report on the signatories’ concerns.

The letter argues that newer AI models can scan networks, write exploit code and chain intrusion steps together without a human operator, a capability known as agentic AI.

That differs from older chatbot tools that only respond when prompted, and it lowers the skill and time needed to breach a hospital network.

From Solo Safeguards To A Joint Industry Front

Individual AI labs had already limited cyber-capable tools on their own before Friday’s letter. Several tightened restrictions on code generation and vulnerability scanning features earlier this year without coordinating publicly.

Also Read: Anthropic Copyright Lawsuit Risk Deepens as Music Giants Sue

This week’s letter marks the issue shifting from scattered company policies toward one sector-wide statement aimed at federal regulators.

Why Washington Is Racing The Clock

The warning lands as U.S. agencies apply the NIST AI risk framework to decide how much autonomy AI systems get near critical infrastructure. In 2021, a human intruder briefly raised sodium hydroxide levels at a Florida water treatment plant in Oldsmar by remotely accessing plant controls.

Signatories argue an AI system capable of finding similar footholds at scale could turn isolated incidents into a coordinated wave.

Whether this threat materializes within the letter’s stated timeline remains unproven, and Congress has not scheduled hearings on the warning.

Read Next: XRP Ledger Quantum Threat Grows After 67M-Fold Cryptography Breakthrough