The UK’s upper chamber wants to rewrite the rules for AI training data, and the government has days to respond.
On March 9, the House of Lords Communications and Digital Committee published a report recommending what legal analysts at Mishcon de Reya describe as a “licensing-first” framework for AI and copyright. According to that analysis, the committee recommends the government rule out a commercial text and data mining exception with a rightsholder opt-out, the approach the UK creative industries have fought against for years.
The committee’s recommendations are advisory. They don’t bind the government. What matters is what comes next: the UK government is required to publish its report and impact assessment by March 18, 2026, per its own commitment. That document will signal whether the licensing-first direction has political support or whether the government charts a different course.
The distinction between these two paths is material for AI developers with UK training data exposure. A licensing framework shifts access to copyrighted content from an uncontested grey zone into a negotiated rights market, with associated cost and compliance obligations. An exception with opt-out keeps the current looser structure, with modifications.
The US settled one piece of its AI copyright question this week, autonomous AI creation doesn’t qualify for protection. The UK is still deciding who gets paid when AI trains on human-created work. Those are related but separate debates, and both are moving at once.