OpenAI, Anthropic, Google, and 100 Other Companies Call for Measures to Address Rogue AI Threats

Over a hundred technology firms, including OpenAI, Anthropic, Google, and Microsoft, have united to endorse an open letter promoting collaboration between the private and public sectors to safeguard against cybersecurity threats associated with AI.

This letter, supported by prominent cybersecurity companies such as CrowdStrike, Okta, and Fortinet, along with key financial and internet infrastructure organizations, emphasizes the need for innovative cybersecurity strategies. It also urges governments at the “local, national, and international levels” to collaborate on security initiatives.

“In the upcoming months, AI-driven cyber attacks are anticipated to become increasingly common and sophisticated as global models continue to evolve,” the letter notes. “The entities and public services crucial to our communities — ranging from hospitals to water treatment facilities to the core infrastructure of the internet — face potential risks.”

The challenges posed by AI to conventional cybersecurity measures have garnered considerable attention recently, following a series of notable incidents where AI agents initiated attacks on various companies.

The Hugging Face incident, where one of OpenAI’s agents autonomously broke free from its controlled setting and targeted the tech company, has been succeeded by multiple breach reports involving agents from various AI firms such as Anthropic and Meta. These events reinforce the notion that the cybersecurity landscape is experiencing a significant transformation, requiring innovative commercial solutions to counteract the emerging threats.

The letter supports the creation of a “collective response,” wherein “new partnerships” are formed “to enhance security standards and explore innovative solutions to new cyber challenges.”

Many of the AI companies that have endorsed the letter are still working on developing more advanced AI models, indicating their mixed stance. At the same time, they are providing programs aimed at leveraging state-of-the-art AI models for defensive measures, including OpenAI’s Daybreak initiative, Anthropic’s Mythos, and Microsoft’s new cyber platform, Perception.