On March 10, 2026, the European Parliament’s plenary adopted the own-initiative report “Copyright and generative artificial intelligence — opportunities and challenges” (procedure 2025/2058(INI), rapporteur Axel Voss), passing 460 in favor, 71 against, with 88 abstentions. The same week, on March 9, Poynter published an advocacy piece by Columbia’s Anya Schiffrin and the European University Institute’s Roberta Carlini arguing for “statutory licensing” — laws that would make AI companies pay for journalism used in training and generation.
Read together, the signal is clear: making AI companies pay for news is graduating from lawsuits and moral appeals into concrete legislative engineering. For any team training or grounding models on web content, this is no longer a distant copyright debate — it is a variable about to enter the cost structure.
The Vote and the Report’s Core Content
The report is not binding legislation, but it is Parliament’s formal position on generative AI and copyright. Its core elements: a “rebuttable presumption” that commercial AI developers used copyrighted works (softened from the committee draft’s irrebuttable version); an extension of press publishers’ and broadcasting rights so rightsholders have “full control” over AI training uses, with “explicit consent” required; recognition that text-and-data-mining exceptions apply to AI training, aligning with the existing EU framework and the AI Act; and exploration of measures against infringing AI outputs that do not block non-infringing content.
Criticism followed immediately. Digital rights group Communia called the report incoherent, offering “no clear path forward,” and warned that its media-content language is broad enough to read press content out of the TDM exceptions — destabilizing the balance of EU copyright law.
What Statutory Licensing Means
The Schiffrin–Carlini push goes further. In a working paper, “How to Update EU and US Copyright Regimes in the Age of AI,” they argue for statutes requiring AI companies to pay publishers for journalism already used in training and for future uses — with pricing that reflects the cost of producing news, brand value, and multiple downstream uses inside AI systems, rather than whatever split individual platforms deign to offer.
Europe, they judge, is the most likely first mover: the EU published two relevant studies in 2025, and Parliament has now voted. Alternative designs include state-administered funds shared among creators and opt-out registries.
The Global Legislative Map
The picture Poynter assembles is sharply divergent:
- Brazil: a draft bill with publisher payments is scheduled for a vote in April
- Indonesia: agreed in a U.S. trade deal not to implement digital taxes or bargaining codes
- South Africa: abandoned its bargaining code, accepting in-kind support such as Meta ad credits
- Australia: discussing a digital levy, after Google and Meta hollowed out its earlier News Media Bargaining Code
- Canada: support for a proposed bargaining code is fading
The litigation front is heating up in parallel. UK publishers formed the SPUR coalition for fair compensation; the Danish publishers’ association is suing OpenAI; EU publishers filed an antitrust complaint against Google in February 2026 over AI news summaries; and a Spanish court fined Meta 524 million euros in November for unfair competition. Danielle Coffey, CEO of the News Media Alliance, which represents roughly 2,000 news organizations, put it bluntly: “If we get the right verdicts, we will have a functional marketplace.”
Why Now
Two forces drive the wave. First, accumulating evidence: AI search and summary products digest news directly on the answer page, visibly eroding publisher traffic and subscriptions, and traditional bargaining mechanisms are losing leverage fast. Second, path dependency: individual licensing deals — Microsoft and Cloudflare have both said they will share AI revenue with publishers — cover only large media; statutory licensing is currently the only mechanism that pulls the long tail of news organizations in.
For AI companies, this is also a cost that can finally be priced. Compared with fragmented litigation risk across jurisdictions, a transparent licensing rate may actually be easier to manage — which is exactly the certainty EU lawmakers hope to supply.
What It Means for Builders
Three practical effects. First, documenting training-data provenance is no longer an academic nicety: a rebuttable presumption of use puts the burden of proof on AI developers, so data pipelines need to answer “which content, obtained when, under which license.” Second, retrieval and summarization products sit in the blast radius — the EU publishers’ antitrust complaint over Google’s AI summaries previews the legal exposure of that product shape, and built-in attribution and licensing design shifts from differentiator to compliance baseline. Third, multi-market products need regional strategies: Brazil, the EU, Australia, and Indonesia are heading to different answers, so content cost models will vary by region — a new input for pricing and market prioritization.
Sources
- A new global push would make AI companies pay for news — Poynter
- Parliament adopts INI report on AI and copyright: no clear path forward — Communia
- European Parliament vote record — HowTheyVote.eu
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.