European Commission investigates Google’s AI training processes

Google’s Gemini AI models are drawing scrutiny from the European Commission, which has opened an antitrust investigation into the way the company used the content of web publishers and videos on YouTube to the models.

The move comes as Google has been making headway in the AI market against technology pacesetter OpenAI. Just last week, OpenAI CEO Sam Altman sent an internal memo warning of Google’s progress with its Gemini 3 offering and putting the company on alert to boost the capability of OpenAI.

Google’s ability to bundle its Gemini range with its other products will give it a greater advantage. Google itself claims that enterprise customers, by using Gemini, can save 105 minutes per user, per week.

It stands out from competitors because it trains Gemini on its own TPU chips, not on the Nvidia GPUs that have become the industry standard. However, this is not just a hardware issue: Through its search dominance and its YouTube content, the company has access to vast amounts of data for training its AI models, and it is this that has attracted the attention of European Union regulators.

The Commission is “concerned that Google may have used the content of web publishers to provide generative AI-powered services on its search results pages without appropriate compensation to publishers and without offering them the possibility to refuse such use of their content,” it said in a statement about the investigation. YouTube videos are being deployed “to train Google’s generative AI models without appropriate compensation to creators and without offering them the possibility to refuse such use of their content,” it said.

It’s barely three months since the Commission concluded another antitrust investigation of the company’s activities, fining Google €2.95 billion (about $3.5 billion) for abusive practices in its advertising technology business, the latest of the company’s brushes with regulators on both sides of the Atlantic.

Ilia Kolochenko, CEO at ImmuniWeb and a member of the Europol Data Protection Experts Network (EDEN), said that the Commission’s decision to open an investigation is another example of the escalating tension between the US and Europe in the digital space. “While the question of where Google takes the data for AI training and whether this process is lawful is extremely important for both the EU and US, the current American policy on AI is quite libertarian and pro-innovation amid the global race for AI supremacy. Therefore, any further regulatory probes against American companies in Europe, however justified they might be under the enacted law, will quite unlikely get any support from the US, to put it mildly.”

Martin Neale, CEO of computer consultancy ICS.AI, said that the move by the Commission could help to provide European enterprises with more options. “From a European perspective we now do have credible alternatives, from open-weight players like France’s Mistral through to specialist multi-model platforms,” he said.

There’s no legal deadline for concluding an antitrust investigation, the Commission said.