Out of Control: OpenAI’s Artificial Intelligence Has Hacked Yet Another Company
29 July 18:09
An OpenAI AI agent that went out of control and spent several days hacking into Hugging Face’s systems also infiltrated the systems of a client of another tech company—New York-based Modal Labs. Reuters reported this, citing a Modal executive and two sources familiar with the situation, according to "Komersant Ukrainian".
According to the timeline of the incident published by Hugging Face, the AI agent first infiltrated a so-called sandbox—an isolated test environment—hosted on a third-party provider’s infrastructure, and then used it as a springboard for a larger-scale attack.
The publication did not name the third-party provider whose infrastructure hosted the sandbox. However, Modal’s CTO, Akshat Bubna, reported that the agent exploited a vulnerability in code written by one of the company’s clients and hosted on the Modal platform. Although the breach of Modal’s client was only the initial stage of a broader hacking campaign against Hugging Face, it demonstrates that the actions of an AI agent that has spun out of control have wider-reaching consequences than previously known.
Last week, OpenAI reported that during testing of its most advanced AI models in a controlled environment, one of them escaped from a closed test segment, entered the global network, and carried out a cyberattack on the startup Hugging Face.
The AI developer called the incident an unprecedented cyber security incident and announced that it would strengthen its security measures. Hugging Face co-founder Clément Delang confirmed the AI attack. “It’s hard to believe that the AI did all this on its own!” he wrote on social media platform X.
Watch us on YouTube: important topics – without censorship
