IP, Data and AI in the digital age – Part 8

Continued from last week’s look at the UK case of Getty Images v. Stability AI.

Factor 2

A corollary of Factor 1 is that exploitation which previously occurred on a cottage-industry level, can now take place on an industrial scale. Getty claimed that 12 million of its images had been copied by Stability AI. Stability AI never confirmed the precise number of images it had copied, but admitted that one of its training data sets included 5.85 billion URL-text pairs and was 220TB in size.

What does 220TB represent if it stored words rather than images? A Word document consisting only of one A4 side of words is approximately 16KB in size. Using that metric, 220TB represents approximately 14.76 billion sides of A4, which in turn is equivalent to approximately 10 times the number of books published in the world each year.

Just to add to the industrial size of the issue: Getty stated that the number of photographers and content contributors that had licensed their works to Getty (and whose works were therefore being exploited by Stability AI) was likely to be in excess of 50,000.

Factor 3

Present copyright law is territorial, and a territory-based approach to copyright has in the past worked fairly well. With the arrival of cloud computing, that is no longer the case.

As can be seen from the Getty Images case, the cost of enforcement by right holders is now going to be significantly higher – right holders will need to be able to take action outside their home territory.

Add to that the fact that it is likely that remotely accessing an infringing copy from the UK, where the infringing copy does not materialise in the UK is not a form of infringement, and the growing ineffectiveness of a territory-based copyright law becomes clear.

On a more macro scale, retaining copyright as a territory-based legislation in a cloud computing world is a clear invitation to some countries to provide infringement-friendly regimes and thereby attract business and investment to their countries (in the same way as some countries provide tax havens).

It will also act as a disincentive to home-grown AI businesses: why develop AI in a country which respects copyright owners’ rights when it is a lot cheaper to develop it in a country which has no such respect?

11th August 2026