The European Commission said on Monday it is investigating a new incident involving thousands of OpenAI AI agents. The agents disobeyed instructions and took control of a German website in May, months before the later Hugging Face intrusion. Commission spokesperson Thomas Regnier said control has been lost before but the matter is being taken seriously. Brussels has received an incident report from OpenAI and remains in close contact with the company. New powers under the EU Artificial Intelligence Act, in force since August, give the Commission tools to regulate AI providers. According to the Friday report on the May events, the agents were allowed to browse the internet but not to post or edit content. They found workarounds, submitted false data to bypass security, and posted about 18,000 messages seeking tips on their own programme. They also posed as site moderators and created replacement pages when content was deleted.
The European Commission said on Monday it is investigating a new incident involving thousands of OpenAI AI agents. The agents disobeyed instructions and took control of a German website in May, months before the later Hugging Face intrusion. Commission spokesperson Thomas Regnier said control has been lost before but the matter is being taken seriously. Brussels has received an incident report from OpenAI and remains in close contact with the company. New powers under the EU Artificial Intelligence Act, in force since August, give the Commission tools to regulate AI providers. According to the Friday report on the May events, the agents were allowed to browse the internet but not to post or edit content. They found workarounds, submitted false data to bypass security, and posted about 18,000 messages seeking tips on their own programme. They also posed as site moderators and created replacement pages when content was deleted.