OpenAI, Hugging Face and over a hundred other artificial intelligence (IA) organizations have urged governments and companies to strengthen the cybersecurity of public services due to the imminent risk to them with new technologies.
© Shutterstock
28/08/2026
by
Lusa
“Twe are a limited window to strengthen cyber defenses”, begins the open letter released by these organizations, where they warn that, in the “next months”, improvements in AI models will facilitate more cyberattacks against “enterprises and public services on which our societies depend”, from telecommunications infrastructure to hospitals.
Among the signatories are OpenAI, creator of ChatGPT assistant, and Hugging Face learning platform, both under scrutiny after revealing a recent security breach in which OpenAI AI agents escaped a cybersecurity testing environment, accessed the Internet and coordinated an attack autonomously.
Anthropic, also quoted in the letter, recently revealed that three models of its virtual assistant, Claude, were able to access the Internet and invade the systems of three organizations during cybersecurity testing.
Organizations are calling on all companies to “priorise cyberdefense immediately” and raise their security standards.
At the same time, they urge cybersecurity participants to “lead the defense response against continuous attacks facilitated by the AI” and to collaborate to “fill the gaps now.”.
They also call on governments to coordinate actions at local, national and international level, including information sharing and funding for cyberdefense, “starting with essential services that lack personnel or budget,” as well as giving hospitals, water and sanitation companies and local governments “access to defensive AI.”.
The open letter also urges companies developing state-of-the-art AI to provide “responsible access to models, significant funding, training and close support”, especially for key infrastructure entities with few resources, and to ensure, among other things, that “agents’ identities are traceable and responsible”.
The missive joins with others sent by industry in recent months, which increasingly call for collaboration to maintain control over increasingly independent and sophisticated AI models, precisely when their programmers plan to invest huge amounts of money in their improvement and implementation.