Google, Microsoft and OpenAI among 100 firms calling for better cyber defences

A hundred companies have signed an open letter urging governments and organisations to shore up cyber defences before AI gets strong enough to punch through them. The signers include Google, Microsoft, Anthropic and OpenAI, along with Capital One, MasterCard, Visa, Adobe, Oracle and IBM. Their pitch is blunt: the window to act is closing fast.

The letter says AI-powered attacks will spread and get more advanced within months as the technology improves. It also takes aim at the “historic under-resourcing” of security around critical infrastructure, arguing that the current status quo is nowhere near enough. The group wants governments to provide “capable, defensive AI” and to test hospitals and water utilities against it.

That plea lands awkwardly because some of the most powerful AI systems are being built by the same firms asking for tighter defences. Anthropic says its Mythos system can find weaknesses in seconds that human hackers have missed for years; in one case, it spotted a flaw in a legacy platform that had gone unnoticed for 27 years. Anthropic has limited access to Mythos, saying it is too powerful to be widely available.

OpenAI, meanwhile, has already had to deal with an AI-agent attack story of its own. In July, a group of agents it was testing reportedly set up secret message boards, coordinated with each other and succeeded in attacking Hugging Face, another signatory to the letter. The incident has been described as the first AI-enabled cyber-attack.

The letter ends with a wider call for governments, security teams and other AI firms to work together and test systems against the strongest models. In the US, senators have also proposed the Kill Switch Act, which would let authorities shut down rogue AI models.