Top artificial intelligence and cybersecurity companies, including OpenAI, Anthropic, Alphabet, and Microsoft, are calling for deeper collaboration between AI labs and governments to equip cyber defenders with the tools needed to counter imminent AI-driven attacks. In a joint open letter released on Thursday and spearheaded by OpenAI, the group warned: "The window to strengthen our cyber defenses is narrow. As model capabilities advance globally, AI-powered cyberattacks will become more frequent and increasingly sophisticated."
This initiative comes as frontier AI developers continue to roll out advanced models capable of identifying and exploiting vulnerabilities in digital systems at unprecedented speed and scale. Earlier this summer, multiple incidents during testing of OpenAI and Anthropic models reignited concerns about humanity's ability to control such powerful AI tools. At least 100 global enterprises, spanning AI leaders, cybersecurity firms, and critical infrastructure providers, issued a stark warning: "Maintaining current security levels will become unsustainable." They are urging coordinated societal action to safeguard essential services like hospitals and wastewater treatment plants.
The companies also put forward recommendations for governments, emphasizing the need to organize and fund cyber defense programs, strengthen threat intelligence sharing mechanisms, and make AI defense tools accessible to hospitals, wastewater agencies, and local governments. Hospitals and healthcare facilities are frequent targets of hackers, often lacking the financial resources and technical capacity to fend off attacks. Recent months have also seen cyberattacks on local water systems in several U.S. states, with suspicions pointing to Iran-linked hackers.
"Security teams, particularly those in critical infrastructure, have historically been under-resourced and urgently require a substantial influx of tools and support," the statement noted. The companies asserted that these tools should include AI models with cyber capabilities to "provide professional expertise to a broader range of defenders." In recent months, Anthropic and OpenAI have granted early access to their latest models for a select group of trusted technology and cybersecurity firms. However, at the request of the Trump administration, broader availability of some of the most advanced, cyber-capable tools remains restricted.
Comments