Opt-in vs. opt-out for AI training

Opt-out regimes let training proceed unless the owner objects — robots.txt conventions, AI-crawler opt-out files, and the EU DSM Article 4 text-and-data-mining exception with its rights-reservation mechanism. Opt-in means a written licence and consent before training happens.

Why it matters to a buyer

Opt-out looks cheaper until diligence. Opt-out signals are contested and unevenly honored, and proving a crawler respected every reservation is hard. Opt-in — a signed licence — is the architecture that survives a legal team’s review.

Why it matters to a data owner

Under EU law you can reserve your rights against text-and-data mining, but a reservation only stops use; it does not pay you. Licensing is the opt-in that does.

Current legal status

The EU DSM Directive’s Article 4 TDM exception permits mining of lawfully accessible works unless rights are expressly reserved, and the mechanics of machine-readable reservation are still being standardized. The EU AI Act now backs this: from August 2, 2025, providers of general-purpose models must maintain a copyright policy that honors Article 4(3) reservations and publish a summary of training content using the Commission’s template. Opt-out conventions such as robots.txt and AI-crawler files carry no uniform legal force of their own and are honored unevenly. Opt-in rests on ordinary contract law and does not depend on any of this settling.

What fiund does about it

fiund is opt-in by construction: nothing is listed without a signed licence granting AI-training rights, and owners approve every use.

Sources

← All rights & provenance guides