More than a hundred companies ask for a window to defend the network

On Thursday, August 27, OpenAI published “A call for collective action on cyber defense.” An open letter. The opening line is this: there is a limited window to strengthen defenses. In the coming months, AI-enabled attacks are expected to become more common and more sophisticated as models around the world get more capable. Hospitals, water treatment plants, the infrastructure that holds up the internet: the letter says those are at risk.

The signatory list is long. Anthropic, Google, Microsoft, Amazon, Cloudflare, banks, insurers, security vendors. Hugging Face signed too. The text does not sell an app. It proposes four cuts: every organization treats defense as a leadership priority; security firms test defenses against what frontier models can already do; governments coordinate and fund, starting with essential services that lack staff and budget; frontier AI companies provide models, money, training, and traceable identity for agents.

The argument in the middle is the “defenders’ window.” The same wave that frightens the attack, the letter says, already gives defense a new way to find old bugs, over-broad permissions, unpatched software. Act now, and the window holds. Wait, and the status quo — short teams, technical debt, weak passwords — will not.

There is an obvious interest, and a reader can see it without a tutorial. The signers also sell the tool the letter wants in defenders’ hands. The BBC and France 24 noted the same cut: the proposed fix runs through the signatories’ own models. The letter asks, in return, for responsible access, authorized testing and verified fixes, not only the speech.

For a reader who only wants to know what is going on, the pointer is this. After a month of news about an agent that left the lab, the industry stopped to write a collective text. This is not the story of one attack. It is the ask for a window, measured in months, with hospitals and water in the first paragraph.