More than 100 companies including OpenAI, Anthropic, Amazon Web Services and Microsoft warned there are only months left to prepare for AI-driven cyberattacks, Axios reported on Aug. 27.

The companies said in an open letter that AI is making sophisticated hacking techniques easy and cheap, exposing critical infrastructure such as hospitals and water treatment plants to hacking threats. "There is not much time left to strengthen cyber defence," they said.

The signatories urged joint action, saying critical infrastructure must be protected in a short period and the cost for hackers using AI tools to infiltrate systems must be raised. The signatories included cloud companies, security companies, AI companies, telecommunications companies, financial firms and think tanks.

The letter said all companies and governments, and security and AI companies, must respond in line with a changing threat environment. It said all organisations should raise internal security standards, fix their most vulnerable points and apply stricter security standards to everything they buy, build and deploy, including AI-generated code.

Security and technology companies should quickly test and deploy AI-based defence tools that critical infrastructure operators can use, and share threat information with each other.

Governments should strengthen channels for sharing practical threat information, coordinate defence at regional, national and international levels, and fund cyber defence.

The letter proposed that frontier AI companies should provide defenders with access to their best-performing response models when major cyber incidents occur and provide funding, training and on-site support to critical infrastructure providers.

The letter comes amid a string of cyberattacks targeting critical infrastructure, including an AI-generated exploit script attack aimed at the U.S. water system. Critical infrastructure such as water utilities and power plants have long had vulnerabilities in facility control tools, but hackers have had to spend significant time understanding these systems' structures. Experts are warning that AI models are significantly reducing the time and effort hackers need to prepare attacks.

Companies participating in the letter did not present specific commitments, deadlines or investment plans.