Google trained its Bard AI with unauthorized content


Since July 24, 2019, the law on related rights has applied in France – a transposition of the European directive on copyright and related rights. Google, however, tends to largely free itself from legislation.

These freedoms earned the Californian firm a new sanction pronounced by the French Competition Authority. She thus faces a fine of 250 million euros. The decision sanctions the non-compliance with the majority of its commitments made in June 2022.

Non-compliance with 4 of its 7 commitments

Google does not dispute this. Gafam has in fact undertaken not to dispute the facts, consequently benefiting from the settlement procedure. In addition, the firm proposed a series of corrective measures aimed at responding to “certain shortcomings identified” by the regulator.

In good faith now? This remains to be confirmed. Because as the Authority points out, this is already the 4th decision rendered in as many years on this same issue of neighboring rights. The emergency measures decreed by the regulator in April 2020 had not been followed.

Consequence for Google: an initial fine of 500 million euros. But the giant is playing for time, while making commitments intended to “put an end to the competition concerns expressed.”

The Accuracy firm appointed to monitor and control their implementation was, however, able to note the lack of cooperation from the leader in online advertising. The latter also failed to respect four of its seven commitments to the press and content publishers.

Bard training: take it or leave it and without remuneration

Far from changing its practices, Google has once again neglected the law on related rights to train its Bard generative AI model. The content of publishers and press agencies was used without information from the rights holders or the Competition Authority.

In addition, Google did not then offer a technical solution allowing content creators to object to their use by Bard without affecting their display on other Gafam services.

Google has thus restricted “the ability of publishers and news agencies to negotiate remuneration.” However, the firm could not ignore the debates throughout 2023 on the question of copyright and the exploitation of web data without consent.

The race to develop generative AI models, in competition in particular with Microsoft and OpenAI, has taken precedence over respect for the law. The company led by Sam Altman is no stranger to this movement.

These practices earned OpenAI a complaint filed by the New York Times. Other authors have also initiated proceedings against the publisher and Microsoft. Open AI now seems ready to make concessions. In March, he signed a partnership with Le Monde.



Source link -97