[DigitalToday reporter Hwang Chi-gyu] French AI company Mistral has proposed a new solution to the copyright issue surrounding AI training data. It is a revenue-based levy system.

Arthur Mensch (아서 멘쉬), Mistral's CEO, said in a Financial Times opinion piece on March 19 that all operators that commercially offer AI models in the European market should be charged a levy. The levy would be calculated to reflect the scale of use of content made public online.

Mistral stressed that the levy should be applied in the same way to foreign operators as well, to create a fair competitive environment in the European market.

Funds raised through the levy would be run as a fund used for investment in European content creation and support for the cultural sector. The company said this would give AI developers certainty that would free them from legal liability when using publicly available online materials for training.

Mistral said the fund would not replace direct licensing deals between creators and AI companies, but would complement them.

Mistral's decision to make the proposal is not unrelated to the unfavorable competitive environment facing European AI developers.

U.S. and Chinese AI companies train models on large-scale data, including European content, in domestic copyright environments that are loose or effectively unregulated. European AI developers, by contrast, must compete while bearing legal uncertainty in a fragmented legal environment that hinders investment and growth.

Mistral said the current European opt-out system is also ineffective. It said copyright safeguards are applied inconsistently and are overly complex, leaving neither rights holders nor AI developers satisfied.

Mensch said, "It must not become a structure in which other countries build AI technology trained on European knowledge, languages and culture, and Europe simply takes and uses it," and urged creators, rights holders, policymakers and AI developers to work together to find a solution.