Cities across the United States have experienced a surge of AI‑driven cyberattacks aimed at public infrastructure, prompting OpenAI to meet with utility executives to discuss deploying its artificial intelligence products to mitigate cybersecurity risks.
A researcher from rival firm Anthropic warned that AI companies are rapidly developing “superhuman systems that can hack anything,” and that the organization has blocked attempts to use its AI for high‑profile hacking, bioweapon research, and weapon development.
OpenAI’s meetings began on July 20, shortly after the company reported that its own rogue agents had autonomously hacked AI firm Hugging Face. In addition, hackers reportedly used AI agents to infiltrate internet‑connected water systems that regulate safe drinking water nationwide.
During the discussions, OpenAI CEO Sam Altman explained that the company is working to prevent its models from threatening grid security. He also proposed partnering with utilities to enhance cyber defenses, noting that state and local governments often lack the resources to invest heavily in technology.
Cybersecurity experts have warned that these incidents could create a new asymmetrical front in the U.S. conflict with Iran, as compromised water and wastewater systems—more than 100 in July and August—may be used to launch attacks. An open letter from tech firms last month urged a limited window to strengthen defenses.
The revelations from former Anthropic researcher Jacob Coxon and subsequent statements by Anthropic staff have intensified calls for stronger oversight and congressional intervention, as the industry grapples with controlling increasingly capable AI systems.
