What is the text and data mining exception?

It exists because copyright law was written for humans reading and copying works, not for software statistically analyzing millions of them at once. Rather than requiring every AI company to negotiate a license before extracting patterns from text, the exception carves out a lane for that kind of automated processing, provided the use fits within what the law defines, which varies by country and often distinguishes research use from commercial use.

It is not the single, uniform shield it is sometimes described as. The EU version lets a rights holder reserve their rights and opt out of the exception for commercial mining specifically; the UK's is narrower. The US has no equivalent statute at all and instead relies on fair use case law, a separate legal doctrine covered under copyright and AI training. A company citing this exception to justify training in one country may have no comparable defense in another.