NEW YORK — What happened
European regulatory bodies issued implementation guidelines under the EU AI Act in late September 2026, establishing strict rules for labeling AI models. Products advertised to businesses or consumers in the region as 'Open Source' must provide open training data documentation, complete source code for data pipeline processing, and unencumbered commercial usage licenses. Models offering downloadable weights under restrictive licenses must be explicitly labeled as 'Source-Available' or 'Restricted Weights.'
Why it matters
Misleading terminology has created compliance confusion for businesses trying to fulfill transparency requirements under consumer protection laws. By establishing legal distinctions between 'Open Source AI' and 'Source-Available AI,' regulators aim to prevent deceptive marketing while granting specific compliance exemptions (such as reduced documentation burdens) to genuine open-source projects created by non-profit entities.
Deep dive
Under the new guidelines, if an enterprise buys or integrates a model labeled 'Open Source,' but the licensor retains the right to revoke commercial access upon reaching a revenue threshold, the licensor risks regulatory fines for misleading commercial practices. This enforcement forces tech marketers to update product pages, repository disclosures, and developer agreements to accurately reflect licensing realities.
Report check
Claims: Regulators claim clear labeling will accelerate enterprise adoption by removing legal uncertainty surrounding license terms. Verification: Legal compliance experts confirm that clear definitions reduce liability risks, though corporate legal teams caution that auditing training data disclosures remains complex.
Open questions
Will US regulatory agencies adopt matching terminology standards, or will global developers face fragmented legal definitions across markets?