Commentary. At Archange Shadows, we believe the central question in the artificial-intelligence copyright debate is not whether AI should exist. It is whether the companies building powerful systems should be allowed to treat the work of journalists, authors and artists as cost-free fuel.

That question has become more urgent as 11 European countries, including Germany, Spain, Poland and Portugal, press the European Commission for stronger rules protecting copyrighted work from AI training. The governments say existing safeguards are difficult for creators to enforce and are urging an easier, more effective opt-out system.[1]

Our position is straightforward: transparency and permission must become the default foundations of commercial AI training. A system that can ingest vast quantities of reporting while disclosing little about what it used—and offering no practical route to compensation—does not represent technological progress alone. It represents a transfer of value from the public sphere to a handful of private companies.

The case for stronger rules

News organizations produce original reporting through expensive, human-intensive work: cultivating sources, checking facts, sending reporters into the field and accepting legal and physical risks. Generative AI systems can summarize, reproduce and compete with that work at enormous scale. If the underlying journalism is copied into training datasets without meaningful consent or payment, publishers face a damaging imbalance: their work helps build products that may reduce traffic, subscriptions and advertising revenue.

The European Parliament has already identified news media as particularly exposed to the displacement effects of generative AI. Its recent policy work has called for stronger transparency and considered additional rights for press publishers and journalists, while also seeking licensing rules that could support media pluralism.[2]

This is not merely a commercial dispute. Independent journalism is civic infrastructure. When reporting disappears, communities lose scrutiny of public officials, corporate power and crises that do not produce immediate profit. Protecting the economic basis of journalism is therefore compatible with protecting democratic discourse.

The counterargument deserves respect

AI developers argue that excessively restrictive copyright rules could entrench the largest technology companies, raise costs for smaller firms and slow research. They also contend that learning from publicly accessible material can resemble the way humans learn from books, articles and images. Some creators may prefer licensing arrangements; others may value broad exposure or want their work used for socially beneficial research.

Those arguments are serious. A blanket prohibition on training from any copyrighted material could make useful tools less accessible and give incumbent rights holders veto power over innovation. Copyright should not become an automatic mechanism for suppressing criticism, experimentation or transformative expression.

But flexibility is not the same as opacity. The industry’s freedom to innovate cannot require everyone else to surrender the ability to know whether their work was used. Nor should “publicly available” be confused with “free for unrestricted commercial extraction.” A newspaper article can be readable online and still embody property, labor and contractual rights.

A workable compromise

Europe should pursue a licensing-first framework with clear exceptions for research and genuinely public-interest uses. AI companies should publish machine-readable records describing the categories and sources of training data, maintain accessible opt-out systems and negotiate collective licenses where individual permissions are impractical. Independent audits should test whether those systems work, and regulators should impose meaningful penalties for evasion.

The European Union’s current approach already allows creators to opt out of certain text-and-data mining, but rights holders and governments say enforcement remains too difficult.[1] The answer is not a symbolic declaration of ownership. It is an operational system that can be used by a small publisher as easily as by a global media conglomerate.

Innovation will survive these rules. What may not survive the absence of rules is the diverse information ecosystem on which innovation, democracy and public trust depend. AI companies have earned the chance to build transformative products. They have not earned the right to build them invisibly on everyone else’s work.

Sources: Euronews, “Germany and 10 other EU countries urge new copyright rules in AI age”; Journal of Intellectual Property Law & Practice, “European Parliament and House of Lords publish reports on AI and copyright”.