In his Opinion on 3 November 2026 in Like Company v Google (C‑250/25), Advocate General Szpunar provides the first major guidance from the CJEU on the relationship between AI chatbots and the press publishers’ right. At its core, the Opinion makes one point clear: what is protected is not the news or the facts themselves, but the specific form in which journalistic content is expressed. This means the legal assessment will depend in particular on how AI systems access press content in real time, process it, and reproduce it in their answers. The risks increase significantly where paywalls or other technical protection measures are circumvented, or where chatbot responses become so comprehensive that they effectively replace the original article. Although the Opinion is not binding, it already sets an important direction for the legal treatment of AI products.
Key takeaways at a glance:
- Facts and news as such remain free, protection only covers the specific journalistic expression.
- In real-time retrieval, the key issues are lawful access to content and whether protected expression is recognisably reproduced.
- Circumventing paywalls and generating answers that effectively replace the original article weigh heavily against AI providers.
- If protected expression remains systematically recognisable in chatbot outputs, this may in principle amount to a communication to the public and therefore an infringement.
- Providers may be liable where such use is inherent in their business model or where they fail to implement credible and effective safeguards.
- AI training may in principle fall under the TDM exception, but the Advocate General addressed this only in passing and left the issue largely open.
Action points for AI providers & deployers:
- Review how their tools access press content in real time and ensure that paywalls and technical protection measures are respected.
- Design outputs to inform users and link to sources rather than replace the original article.
- Implement and document guardrails against systematic reproduction and test outputs regularly.
- Reassess training data practices and opt-out mechanisms in light of the TDM exception.
Action points for publishers:
- Focus enforcement on the recognisable reproduction of protected expression rather than on information as such.
- Ensure that technical protection measures are effective and properly documented.
- Use the Opinion as leverage in licensing discussions with AI providers.
Why this matters:
Guardrails, output design, and lawful access to content are becoming core legal risk factors for AI products. While the Opinion is not binding, it sets the tone ahead of the CJEU’s judgment and is likely to carry significant weight.
