Alphabet-owned Google has been handed a €250 million fine for breaches of EU intellectual property rules over the use of content to train its AI platform.

The Autorité de la concurrence on Wednesday said that Bard – which has since has been rebranded as Gemini – was trained on content from publishers and news agencies without notifying them.

The fine stems from a copyright dispute in the country in a case that was triggered by complaints from local news organisations.

Google, which has said it would not contest the facts as part of settlement proceedings, had dropped its appeal against an initial €500 million fine issued to the company in 2022, but the watchdog had reopened the case over the AI program.

In a statement issued on Wednesday, the regulator said that Google had violated the terms of four out of the seven commitments to which it had agreed in the settlement. These terms included conducting negotiations with publishers in good faith and providing transparent information on how it sourced its content.

The regulator said that Bard, launched in 2023, was trained on data from unspecified outlets and agencies without consent or informing them. This, the regulator said, hindered the ability of publishers to negotiate fair prices.

It said: "Subsequently, Google linked the use of the content concerned by its artificial intelligence service to the display of protected content.”

The issue of intellectual property rights infringement to train AI models has been brought to the fore over the past 12 months. The New York Times currently has a lawsuit against OpenAI, while the Microsoft-backed company earlier this year admitted to the UK parliament that “it would be impossible to train today’s leading AI models without using copyrighted materials.”

---