Permission to appeal granted in Getty Images v Stability copyright claim

[Oliver Fairhurst]

Permission to appeal granted in Getty Images v Stability copyright claim

Oliver Fairhurst Wednesday, January 07, 2026 - AI, Getty Images v Stability AI, Mrs Justice Joanna Smith DBE, Oliver Fairhurst, secondary copyright infringement

Readers will be well aware of the outcome of the Getty Images v Stability AI case, in which Getty's claim failed on secondary copyright infringement (Getty having abandoned its case on primary copyright infringement), but succeeded on a limited element of its trade mark claim (in relation to historic models reproducing Getty watermarks). See our IPKat post here.  

Given the novelty of the case, and the stakes, it is unsurprising that Getty and Stability both sought to appeal the findings on which they lost. The judge, Joanna Smith DBE heard the application in December for permission to appeal following that judgment, and granted Getty permission to appeal the rejection of its secondary infringement claim (Getty Images (US) Inc & Ors v Stability AI Ltd (Re Form of Order) [2025] EWHC 3343 (Ch)). She refused Stability's application to appeal the trade mark infringement finding, meaning that Stability will need to go directly to the Court of Appeal to seek permission. 

Appeal on secondary copyright infringement

This was the most interesting part of the case to many observers.  The crux of Getty's argument was that Stable Diffusion was an "infringing copy" of the works on which it was trained, being an "article" whose making would, had it occurred in the UK, have infringed UK copyright. 

The judge rejected the claim on the basis that the words "infringing copy" must mean that the "article" (which she accepted Stable Diffusion was) must actually contain the works on which it was trained. Partly based on evidence, and partly on concessions made by Getty, she decided that it did not contain those works. This finding has been the subject of intense criticism by critics of LLMs and the process of their creation, and praised as correct by proponents of LLMs. 

This element of the case could, if Getty was successful, or is on appeal, could be (to put it mildly) impactful. It could mean that LLMs trained without authorisation or other lawful excuse under UK copyright law, could not be possessed, sold, hired, advertised, or distributed in the course of business in the UK. 

The judge held that the appeal on this point does have a real prospect of success, and concerns a "concerns a pure question of law, namely a matter of statutory construction on which the minds of reasonable lawyers may differ."

A complicating factor is that the issue did not, at trial, to this Kat's mind, concern a pure question of law. It concerns a very complex set of factual findings and concessions by Getty (see [552]-[560]). Assuming those concessions were not made, and the factual findings are not overturned (particularly as to the relevance of memorisation, the question on appeal as to whether the "infringing copy" must actually contain the works will nevertheless be fascinating.

Appeal on trade mark infringement

Stability sought permission to appeal the finding against it that it had infringed Getty's trade marks. This Kat has not seen the grounds of appeal (but would welcome a copy if any of the legal teams involved would be willing to share them). 

It appears from the judgment that there were three "main" grounds of appeal, which the judge found had no real prospect of success. The first two grounds were that the findings were rationally unsupportable, which is always a high bar and one that the judge will rarely accept. 

The refusal of permission on the third ground warranted more explanation. This ground focused on the identification of the average consumer in the particular context of watermarks. There were several sub-grounds of appeal (not all addressed here):

• Stability argued that the judge ignored the two purposes of watermarks (i.e. to make the images unusable and to identify the owner to allow licensing), and that as most users would discard a watermarked image, so were not acting as a consumer. This was refused on the basis that there was no evidence at trial on this point, and there was no explanation of why the consumer would not be responding to both purposes of the trade mark at the same time.  
• Stability also argued that it was illogical to consider that the consumer would perceive the watermark as an indication of trade origin, which the judge rejected on the basis on the lack of evidence and the attempt to attribute a higher level of sophistication to the average consumer than was appropriate. 

This Kat thinks that the trade mark findings at first instance might have a significant effect on the way that LLMs operate in at least the UK. The finding that AI deployers use trade marks in their own commercial communication might give hope to publishers in particular if copyright does not provide protection for their content. 

Do you want to reuse the IPKat content? Please refer to our 'Policies' section. If you have any queries or requests for permission, please get in touch with the IPKat team.