What happened
On 28 October 2026 the US Copyright Office published a notice of proposed rulemaking (91 FR 67788; FR Doc. 2026-21939) for the tenth triennial DMCA §1201 anti-circumvention rulemaking. It intends to recommend renewal of all but one existing exemption and opens three rounds of public comment on newly proposed exemptions. Notably, responding to the Association of American Publishers, the Office states that 'circumventing TPMs to compile AI model training data sets is permitted by the exemption' — confirming that the existing text-and-data-mining exemption for nonprofit academic researchers covers compiling AI training data. Comment deadlines: initial comments 7 December 2026, opposition 22 February 2027, replies 22 March 2027.
Why it matters
This is the authoritative US Copyright Office action clarifying the legal reach of the AI-training-data mining exemption under the DMCA, and sets the framework for the next three years of copyright-access exemptions that materially affect AI training and research. It also signals regulators will lean against publishers seeking to fence off AI training use of copyrighted works under the existing nonprofit exemption.
Action needed
Researchers, developers and rights holders engaged in AI training-data compilation should review the proposed exemption text and comment by 7 December 2026; the confirmation narrows a licensing/takedown lever publishers might otherwise assert against nonprofit AI training.