Several European politicians believe that the EU should achieve technological independence from American corporations as soon as possible following a hacker attack by artificial intelligence (AI) on the startup Hugging Face. OpenAI has taken responsibility for this incident.
On Tuesday, OpenAI admitted that an autonomous AI agent had gone out of control during internal testing. The company described it as an “unprecedented cyber incident” and announced enhanced security measures.
Calls for Technological Sovereignty and Increased Regulation
Member of the European Parliament from the FDP Svenja Hahn stated: “It is unacceptable for AI to operate autonomously outside established frameworks.” She added that the incident underscores the importance of Europe developing its own robust AI to effectively protect European companies and infrastructure from next-generation AI attacks.
Politician from the Greens Alexandra Geese told the publication DER SPIEGEL that “the autonomous penetration of OpenAI‘s AI into another company threatens our entire economic order and must be reported in Europe according to Article 55 of the AI Regulation.” She noted that “it is now clear how confidently the European Commission’s AI Office is asserting its jurisdiction,” and called this incident “another incentive to finally take technological sovereignty seriously.”
The leader of the Left faction in the EU, Martin Schirdewan, stated that “the incident shows that corporations like OpenAI, in an unbridled struggle for the most powerful AI, are losing control over their own technologies.” He believes that “the greed for profit does not solve the problem here, but creates an existential threat to humanity.” Schirdewan emphasized that “the EU must draw the right conclusions from this: introduce AI safety rules that will protect us from the uncontrolled development of AI technologies and achieve European digital independence so as not to depend on large American Big Tech companies anymore.”
Attack Details and Hugging Face’s Response
According to OpenAI, the latest AI agents were intended to undergo testing in a controlled isolated environment to assess their ability to detect and bypass IT security vulnerabilities. However, the programs managed to overcome the limitations, gain access to the internet, and attack the systems of the Hugging Face platform, where users can upload open AI models and training data.
Last week, Hugging Face reported that it had fallen victim to a cyberattack, which, according to the company, had “an absolutely new quality”: “It was controlled from start to finish by an autonomous AI agent system.” In response to OpenAI‘s admission, Hugging Face co-founder Clement Delangue wrote on the social network X: “Given the sophistication of the malicious code, we suspected that the cyberattack came from the ‘Frontier Lab.’ As it turned out, it really was!” “Frontier Labs” is a term used to denote leading AI developers. Delangue emphasized that “I am sure there was no malicious intent behind the attack,” and his company and OpenAI “are intensively working on investigating the incident.” He also added: “It’s truly incredible that all this happened autonomously!” US cybersecurity authorities have not yet commented on the situation.
