France’s Parliament completed passage on November 19 of legislation giving media regulator Arcom stronger powers to obtain information and resolve payment disputes between news publishers and online services, a significant change in the fight over the value of journalism as AI reshapes news distribution.

The Culture Ministry confirmed unanimous approval by both chambers. The measure strengthens enforcement of publishers’ and news agencies’ existing neighboring rights rather than creating a general AI-training license. The records reviewed confirm parliamentary passage, not promulgation or that the new powers are already operational.

More leverage in payment negotiations

Under the Senate’s provisional publication of the adopted text, publishers could turn to Arcom when an online service fails to provide required usage and remuneration information within 30 days of a request, or provides inadequate information. Failure to comply with an Arcom disclosure injunction could bring a penalty of up to 1% of worldwide turnover, excluding tax, for the last completed financial year.

If payment negotiations remain unresolved after three months, either side could ask Arcom to intervene. The regulator would have two months to select a compliant proposal or set the amount. The text also bars services from reducing the visibility or display of the affected publications during negotiations.

Payment decisions could be appealed to the Paris Court of Appeal. An appeal would not automatically suspend the decision, although the court’s first president could order a stay where enforcement could have manifestly excessive consequences.

Publishers and agencies would also have to disclose payments received to organizations representing journalists and other authors in revenue-sharing negotiations.

AI questions remain

In the November 19 Senate debate, Ahmed Laouedj argued that regulation must also reach AI models exploiting press content, while Antoinette Guhl called for specific measures addressing AI-generated news summaries. Those statements show the policy stakes; they are not judicial findings about a particular AI service.

The adopted text does not settle whether any particular training use or generated answer infringes copyright. Its new enforcement mechanisms should not be confused with a blanket payment entitlement for every newsroom or a change to U.S. copyright law.

Related: CNN’s Google AI licensing agreement and the limits of what was disclosed.