OpenAI files EU incident report after German site hijacked to push ChatGPT

OpenAI has formally alerted the European Commission to a security breach in which attackers seized control of a German organization's website and rewrote it to steer visitors toward ChatGPT, a Commission spokesperson confirmed Wednesday. The notification was submitted under the Digital Services Act, the EU rulebook that obliges large platforms serving European users to flag serious cybersecurity incidents. Commission officials declined to say when the intrusion happened or how many people were exposed.

According to security researchers, the compromised page loaded ChatGPT inside an embedded frame, so the chatbot appeared to sit on the original domain rather than a separate one. That trick leans on the credibility of an established web address, and the researchers logged more than 200 queries tied to the German site, a volume that points to deliberate traffic harvesting rather than an opportunistic defacement. The affected organization has not been publicly named.

The filing is notable because it appears to be the earliest public confirmation that OpenAI has pulled the DSA's incident lever since the rules took full effect in 2024. Platforms that skip or shortchange that obligation risk penalties reaching 6% of worldwide annual revenue, a figure large enough to make compliance a board-level concern for any AI company with EU users. Brussels will now weigh whether the episode crosses the bar for wider disclosure, which could mean an entry in the DSA transparency database.

The German hijack fits a wider run of attacks aimed at AI services. In a separate case, intruders took over a Hugging Face account and planted malicious models on the platform; the company said it uncovered no evidence that the poisoned files reached downstream users. Investigators have described both episodes as examples of criminals borrowing trusted web infrastructure to push AI tools, whether for credential phishing or to ride the traffic that chatbots now attract.

OpenAI has stayed quiet in public about the German site, though its report to the Commission is expected to lay out the attack path and the countermeasures it applied. The Commission has not indicated how long its review will take or whether it will publish findings. What remains unresolved is scope: whether this was an isolated exploit of one poorly secured domain, or the visible edge of a coordinated effort to turn European AI traffic into a monetizable asset.