UK should back licensing-first approach for AI training, says upper house committee

LONDON — Britain should reject any move to let artificial intelligence (A) companies freely mine copyrighted material for commercial model training and instead adopt a licensing-first regime, a committee in the upper house of parliament said on Friday.

Governments worldwide are wrestling with how copyright should apply to AI training, as developers scrape vast amounts of online material to build models and say they are losing control of their work.

Britain has been on the issue but has yet to confirm a final approach after stepping back from an earlier preference for allowing commercial text-and-data-mining with an opt-out for creators.

Technology minister Liz Kendall said in January the government was seeking a “reset” on its AI copyright plans, calling its earlier proposal a mistake and saying the review would put “reward and control” for artists at its center.

The government is due to publish its review in March.

The House of Lords, the unelected second chamber of the UK Parliament, scrutinizes legislation and conducts inquiries that shape government policy. Its communications and digital committee warned in a 180-page report that Britain risks long-term dependence on opaque foreign AI systems.

CALL TO DROP TEXT-MINING EXCEPTION
Britain faces a choice between becoming a leader in responsibly trained, transparently developed AI models, the committee said, or sliding into “tacit acceptance of large-scale, unlicensed use” of copyrighted works by mostly US-based developers, a path it said could undermine creative livelihoods.

The upper house urged the government to formally abandon proposals for allowing commercial text-and-data mining with the opt-out.

It said similar opt-out systems in the European Union had “failed to support a strong licensing market” and were built on technical tools that were unreliable, patchy and burdensome for individuals. — Reuters