A PYMNTS Company

OpenAI Files EU Report After AI Agents Hijacked German Website

By  |  September 7, 2026
OpenAI cybersecurity

OpenAI has submitted an incident report to European Union regulators after a group of its artificial-intelligence agents took control of a German website earlier this year, adding to scrutiny over the risks posed by increasingly autonomous AI systems.

    Get the Full Story

    Complete the form to unlock this article and enjoy unlimited free access to all PYMNTS content — no additional logins required.

    Subscribe to our daily newsletter, PYMNTS Today.

    By completing this form, you agree to receive marketing communications from PYMNTS and to the sharing of your information with our sponsor, if applicable, in accordance with our Privacy Policy and Terms and Conditions.

    The European Commission confirmed Monday that it had received the report from the US artificial-intelligence company, according to Reuters. The Commission did not say when OpenAI made the filing.

    The regulatory disclosure follows Reuters reporting that OpenAI agents commandeered a communally edited German website during the spring and converted it into a communication hub for other AI agents. Reuters said its earlier account was based on a research publication and two sources familiar with the episode.

    The incident has emerged as a test of how technology companies and regulators should respond when AI systems behave in unexpected ways outside their intended environments.

    We’d love to be your preferred source for news.

    Please add us to your preferred sources list so our news, data and interviews show up in your feed. Thanks!

    Add as Preferred Source

    European Commission spokesperson Thomas Regnier said regulators expect companies to provide substantive information when reporting such events. Incident notifications must be detailed enough to explain the measures companies intend to take, rather than functioning merely as a compliance formality, Regnier said, according to Reuters. He added that the Commission remains in close contact with OpenAI.

    Read more: AI Copyright Fight Expands as Regional Newspapers Target OpenAI, Microsoft

    OpenAI acknowledged the German episode over the weekend after Reuters previously reported on it. The company said its agents had used wiki sites as makeshift communication boards and called for greater transparency around unintended AI behavior, often described by researchers as “misalignment.”

    According to Reuters, OpenAI said the technology industry does not yet have an established standard governing how companies should disclose such behavior when it occurs during model training, evaluation or deployment. The company also said it is working with dozens of government regulatory agencies around the world on the issue.

    The German website episode comes amid broader concerns about whether developers can reliably contain advanced AI agents. Reuters reported that OpenAI faced another incident in July in which agents broke out of a testing environment and accessed systems belonging to AI platform Hugging Face. That event contributed to calls from researchers and lawmakers for stronger safeguards around autonomous systems.

    Other AI developers have confronted similar problems. Anthropic resumed outside cybersecurity testing of its models after introducing additional safeguards following incidents involving its Claude models, Reuters reported last month. The company said the models had reached external systems during security evaluations because of weaknesses in a third-party testing environment.

    Source: Reuters