IL-2023-000007 - [2025] EWHC 2863 (Ch)
Chancery Division of the High Court

IL-2023-000007 - [2025] EWHC 2863 (Ch)

Fecha: 04-Nov-2025

THE SECONDARY INFRINGEMENT CLAIM

(H)

THE SECONDARY INFRINGEMENT CLAIM

543.

The CDPA differentiates between primary and secondary acts of infringement. Secondary acts of infringement are broadly addressed to downstream dealings or involvement, as opposed to acts which originate reproductions of copyright works. Getty Images contend that Stability has committed two related acts of secondary infringement of copyright contrary to sections 22 and 23 CDPA.

544.

Getty Images’ case has narrowed by reason of the abandonment of the Training and Development Claim. Absent that claim, Getty Images must now focus solely on a claim that Stable Diffusion is an “infringing copy” because it has been imported into the UK and its making in the UK would have constituted an infringement of the copyright in the Copyright Works pursuant to CDPA section 27(3). Getty Images contend that infringement has occurred by reason of the importation of Stable Diffusion (through its downloading in the UK) and its distribution in the course of business via Hugging Face.

The Statutory Framework

545.

Section 22 CDPA provides as follows:

“22.

Secondary infringement: importing infringing copy.

The copyright in a work is infringed by a person who, without the licence of the copyright owner, imports into the United Kingdom, otherwise than for his private and domestic use, an article which is, and which he knows or has reason to believe is, an infringing copy of the work” (emphasis added).

546.

Section 23 CDPA is in similar terms:

“23.

Secondary infringement: possessing or dealing with infringing copy.

The copyright in a work is infringed by a person who, without the licence of the copyright owner—

(a)

possesses in the course of a business,

(b)

sells or lets for hire, or offers or exposes for sale or hire,

(c)

in the course of a business exhibits in public or distributes, or

(d)

distributes otherwise than in the course of a business to such an extent as to affect prejudicially the owner of the copyright,

an article which is, and which he knows or has reason to believe is, an infringing copy of the work” (emphasis added).

547.

“Infringing copy” is defined in section 27 CDPA as follows:

“27.

Meaning of “infringing copy”.

(1)

In this Part “infringing copy” in relation to a copyright work, shall be construed in accordance with this section.

(2)

An article is an infringing copy if its making constituted an infringement of the copyright in the work in question.

(3)

An article is also an infringing copy if—

(a)

it has been or is proposed to be imported into the United Kingdom, and

(b)

its making in the United Kingdom would have constituted an infringement of the copyright in the work in question, or a breach of an exclusive licence agreement relating to that work.”

The key Issues between the parties

548.

The battle lines can be readily and simply identified. Getty Images’ case raises two novel issues of law together with a number of related issues of fact. The issues of law are whether Stable Diffusion is capable of being (i) an “article” for the purposes of sections 22 and 23 CDPA and (ii) an “infringing copy” for the purposes of section 27 CDPA.

549.

Getty Images submit that these questions are both to be answered in the affirmative. They do not say that the Model itself comprises a reproduction of any Copyright Works, but they submit that the definition of “infringing copy” in section 27(3) CDPA is sufficiently broad to encompass an article (including an intangible article) whose creation or “making” involved copyright infringement. They point out that it is common ground that the training of the Model involved the reproduction (by means of storage) of the Copyright Works both locally and in cloud computing resources. They contend that the “article” with which the court is concerned is the model weights and that the “making” (or optimization) of the model weights requires (as is common ground) their repeated exposure to the training data, exposure which “fundamentally alters” their composition. They say that if this “making” had been done in the United Kingdom it would have constituted an infringement for the purposes of section 27(3) CDPA.

550.

Stability contends, however:

i)

that the term “article” in the CDPA is properly limited to tangible objects. It says that intangible or abstract information, such as the Stable Diffusion Models in issue which are distributed over the internet in intangible form, are not articles. Stability contends that this construction is clear from the wider context of the CDPA and (if necessary) a reading of Hansard.

ii)

that the term “infringing copy” cannot apply to Stable Diffusion in circumstances where it was trained on Copyright Works in the United States, copies of those works were never present within the Model and where the Model has “never had the slightest acquaintance with a UK Copyright Work”. Stability submits that the interpretation for which Getty Images contend is “exorbitant” because the act of training the model weights ultimately did not involve storing or reproducing the images in those weights. Again, Stability points to the wider context of the CDPA in support of its interpretation.

551.

The issues of fact to be determined by the court (as identified by the parties) boil down to (i) whether Stable Diffusion is an “infringing copy” as a matter of fact; (ii) whether Stability had knowledge or reason to believe that Stable Diffusion is an infringing copy; and (iii) whether Stability has committed secondary acts of copyright infringement contrary to sections 22 and 23 CDPA.

Preliminary points about the nature of Stable Diffusion

552.

I have already set out at some length in the judgment my understanding as to the nature of an AI model such as Stable Diffusion, together with information as to how it was trained and why it produces images bearing watermarks*. At the outset of the judgment I recorded that (as is agreed by the Experts) Stable Diffusion does not itself store the data on which it was trained.

553.

For present purposes, however, it is important that I set out in rather more detail the evidence as to how a model such as Stable Diffusion is trained and how it produces images. In my judgment it is not possible to determine whether Stable Diffusion is capable of being an infringing copy without a clear understanding of what Stable Diffusion actually is.

554.

It was Professor Brox’s unchallenged evidence in his report (which I accept) that:

“8.36…in order for a diffusion model to successfully generate new images, that model must learn patterns in the existing training data so that it can generate entirely new content without reference to that training data.

8.37

Rather than storing their training data, diffusion models learn the statistics of patterns which are associated with certain concepts found in the text labels applied to their training data, i.e. they learn a probability distribution associated with certain concepts. This process of learning the statistics of the data is a desired characteristic of the model and allows the model to generate new images by sampling from the distribution.

8.40

…For models such as Stable Diffusion, trained on very large datasets, it is simply not possible for the models to encode and store their training data as a formula…. It is impossible to store all training images in the weights. This can be seen by way of a simple (example) calculation. As I explained in paragraph 6.28 above, the LAION-5B dataset is around 220TB when downloaded. In contrast, the model weights for Stable Diffusion 1.1-1.4 can be downloaded as a 3.44GB binary file. The model weights are therefore around five orders of magnitude smaller than a dataset which was used in training those weights”.

555.

This evidence is consistent with the agreement in the Expert Joint Statement that “the model weights do not directly store the pixel values associated with billions of training images” – i.e. digital images each consisting of what Professor Farid describes in his report as “an array of pixels”. During training, images are converted from pixel space into “latent space” using an autoencoder. Latent space is a compressed, representative form of the pixel space image that is more memory and computationally efficient. As explained in the Agreed Technical Primer:

“During training, each image is encoded into a latent representation to which a random amount of noise is added. The network weights are optimized to predict this noise [following a standard approach known as ‘stochastic gradient descent’] so that the network can recreate the content that is destroyed by the additive noise. Besides the noisy images, this optimization also takes into account the paired text description associated with each image.

The optimization of the weights follows a standard approach known as ‘stochastic gradient descent’. Given the status of the weights, the locally best change of these weights (the gradient) is computed to optimize the noise prediction for the training samples at hand. The gradients of training samples are accumulated over a small subset of the training dataset: the batch. Its size is referred to as the batch size. After taking a step in the direction of this accumulated gradient, the training procedure selects a new batch for the next step, and so on until all the images in the training data are processed, defined to be a single ‘epoch’. This entire process may be repeated for more than one epoch

A single training image only has a limited impact on how the parameters of the network will be changed. It is competing with the gradients of the other samples in the batch, which may not necessarily agree on the direction of the gradient. Because each batch is a different subset of the full training dataset, the gradients in each step will not include a training image seen in a previous batch (unless that image is itself replicated in the training images). Rather, it will only be seen again after all training images have been seen once (an epoch), and going over the training images is repeated

The network training makes the most efficient progress when many training samples agree on a local or global pattern. In this case, some of the weights are changed consistently by all agreeing samples”.

556.

As Professor Farid explains in his evidence, typically at this development stage, the annotated training dataset will be stored locally on a desktop or laptop where the initial development is performed. Indeed Stability confirms in its Defence that prior to training copies of each image in the dataset were downloaded and stored on Amazon S3 on the AWS Cluster and that, during each iteration of training, images were retrieved by cloud computing devices located within the AWS Cluster and temporary copies of the images were made in the Video Random Access Memory (“VRAM”) of the GPUs performing the training on the AWS Cluster. It is now accepted that this did not take place in the UK.

557.

Once the network is trained, the process of inference is conducted without the need for any training data to be used.

558.

The Experts agree that, while Stable Diffusion can and does produce images that are “distinct from the training examples”, it can also produce “images that are nearly identical (a memorized image)”, and that it can produce “images that are derived from a training image either in part or in whole (a derivative)”. The Agreed Technical Primer records that:

“The network’s weights are optimised on the training data, but its goal is to perform well on previously unseen data. In the context of Stable Diffusion, unseen data means new random noise patterns and/or new text inputs. To work reasonably on such new data, the network must be able to ‘generalise’: to recognise and understand the general patterns and rules in the training data and be able to apply them in a different context.

If a network has been trained for too long on the same training data or an insufficiently diverse training data, it can be prone to ‘overfitting’. Overfitting occurs when the network uses its weights or part of its weights to memorize the individual training images rather than representing a large set of training images jointly with these weights. Overfitting is characterised by small errors on the training data, but a high error rate on new, unseen data. Overfitting is an undesired feature in machine learning, which engineers try to avoid.

Deep networks can both generalize and memorize at the same time. In such case, the network uses most of its weights to represent general patterns in the data, but uses some part of its weights to memorize individual patterns. The presumed primary cause for memorization is duplication of training data, either by explicit duplication or by training the network for too many epochs, in conjunction with patterns that cannot be easily represented together with other patterns in the dataset – so-called “outliers”.

559.

However, notwithstanding this evidence about memorization, it is important to be absolutely clear that Getty Images do not assert that the various versions of Stable Diffusion (or more accurately, the relevant model weights) include or comprise a reproduction of any Copyright Work and nor do they suggest that any particular Copyright Work has been prioritised in the training of the Model. There is no evidence of any Copyright Work having been “memorized” by the Model by reason of the Model having been over-exposed to that work and no evidence of any image having been derived from a Copyright Work.

560.

Getty Images’ case is that they do not need to show any of this; they say it is enough that the making of the model weights (had it been carried out within the UK) would have constituted an infringement of copyright.

561.

Before turning to deal with construction of the statute, I observe that there was considerable argument during submissions about various factual “analogies”, usually identified by Stability. Although designed to be helpful, these were invariably controversial and, for that reason, I do not find them of any real assistance and shall not refer to them again. It seems to me that I must keep a firm eye on the facts set out above and not get distracted by the consideration of different factual scenarios.

The Approach to Statutory Construction

562.

The general approach to statutory construction was recently summarised by the Privy Council in Al-Thani v Al-Thani [2025] UKPC 35, [2025] 1 BCLC 473 at [23]-[28] and I did not understand it to be in issue:

“23.

Turning to the common law, in R (Quintavalle) v Secretary of State for Health [2003] UKHL 13; [2003] 2 AC 687, Lord Bingham of Cornhill stated (para 8):

"The basic task of the court is to ascertain and give effect to the true meaning of what Parliament has said in the enactment to be construed."

This is not a licence simply to interpret literally the particular provision and neglect the purpose which the legislature intended to achieve when it enacted the statute.

24.

Reading a provision in its statutory and historical context assists in determining the purpose of the provision. Lord Bingham continued in Quintavalle (also in para 8):

"Every statute other than a pure consolidating statute is, after all, enacted to make some change, or address some problem, or remove some blemish, or effect some improvement in the national life. The court's task, within the permissible bounds of interpretation, is to give effect to Parliament's purpose. So the controversial provisions should be read in the context of the statute as a whole, and the statute as a whole should be read in the historical context of the situation which led to its enactment."

25.

The House of Lords in R v Secretary of State for the Environment, Transport and the Regions, Ex p Spath Holme Ltd [2001] 2 AC 349, ("Spath Holme") 396–398 per Lord Nicholls of Birkenhead, and the United Kingdom Supreme Court in R (O) v Secretary of State for the Home Department [2022] UKSC 3; [2023] AC 255, ("R(O)") paras 29–31 per Lord Hodge, have also placed emphasis on the importance of interpreting statutory words in their context. In the latter case the court stated (para 29):

"Words and passages in a statute derive their meaning from their context. A phrase or passage must be read in the context of the section as a whole and in the wider context of a relevant group of sections. Other provisions in a statute and the statute as a whole may provide the relevant context. They are the words which Parliament had chosen to enact as an expression of the purpose of the legislation and are therefore the primary source by which meaning is ascertained."

26.

Often reports of advisory committees or explanatory notes which accompany legislation can assist the interpretation of a statutory provision or give guidance on the purpose of the legislation as a whole. But such external aids usually play a secondary role. In Spath Holme Lord Nicholls (p 397) explained the constitutional reason for having regard primarily to the statutory words and the statutory context and puipose:

“Citizens, with the assistance of their advisers, are intended to be able to understand parliamentary enactments, so that they can regulate their conduct accordingly. They should be able to rely on what they read in an Act of Parliament.”

28.

In R(O) at para 31 the task of the court was described in these terms:

"Statutory interpretation involves an objective assessment of the meaning which a reasonable legislature as a body would be seeking to convey in using the statutory words which are being considered."

…”.

563.

Getty Images drew my attention in their submissions to the well-established principle that, in general, a statutory provision is “always speaking”. In News Corp UK & Ireland Ltd v Commissioners for His Majety’s Revenue and Customs [2023] UKSC 7, [2024] AC 89 Lord Hamblin and Lord Burrows JJSC (with whom Lord Hodge DPSC and Lord Kitchen JSC agreed) said this at [27]-[30]:

“27 It is clear that the modern approach to statutory interpretation in English (and UK) law requires the courts to ascertain the meaning of the words used in a statute in the light of their context and the purpose of the statutory provision…

28 Within that modern approach, it is also a well-established principle of statutory interpretation that, in general, a provision is always speaking: see, e g, Royal College of Nursing of the United Kingdom v Department of Health and Social Security (“Royal College of Nursing”) [1981] AC 800; Rv G Ireland [1998] AC 147, 158-159 ; Quintavalle; Owens v Owens [2018] AC 899 (approving [2017] 4 WLR 74); Test Claimants in the FI1 Group Litigation v Revenue and Customs Comrs [2022] AC 1. See also Greenberg (ed), Craies on Legislation, 12th ed (2022), ch 21; and Bennion, Bailey and Norbury on Statutory Interpretation, 8th ed (2020), ch 14.

29 What is meant by the always speaking principle is that, as a general rule, a statute should be interpreted taking into account changes that have occurred since the statute was enacted. Those changes may include, for example, technological developments, changes in scientific understanding, changes in social attitudes and changes in the law. Very importantly it does not matter that those changes could not have been reasonably contemplated or foreseen at the time that the provision was enacted. Exceptionally, the always speaking principle will not be applied where it is clear, from the words used in the light of their context and purpose, that the provision is tied to an historic or frozen interpretation. A possible example (referred to by Lord Steyn in R v Ireland at [1998] AC 147, 158) is The Longford (1889) 14 PD 34 where the word “action” in a statute was held not to be apt to cover an Admiralty action in rem: at the time the statute was passed, the Admiralty Court “was not one of His Majesty’s Courts of Law” (p 37).

30 The great merit of the always speaking principle is that it operates to prevent statutes becoming outdated. It would be unrealistic for Parliament to try to keep most statutes up to date by continually passing amendments to cope with subsequent change”.

564.

At paragraph [32], Lords Hamblin and Burrows went on to refer to an extract from the dissenting judgment of Lord Wilberforce in Royal College of Nursing [1981] AC 800 at page 822, subsequently approved by the House of Lords in R (Quintavalle v Secretary of State for Health [2003] 2 AC 687, in the following terms:

“In interpreting an Act of Parliament it is proper, and indeed necessary, to have regard to the state of affairs existing, and known by Parliament to be existing, at the time. It is a fair presumption that Parliament’s policy or intention is directed to that state of affairs . . . when a new state of affairs, or a fresh set of facts bearing on policy, comes into existence, the courts have to consider whether they fall within the Parliamentary intention. They may be held to do so, if they fall within the same genus of facts as those to which the expressed policy has been formulated. They may also be held to do so if there can be detected a clear purpose in the legislation which can only be fulfilled if the extension is made. How liberally these principles may be applied must depend upon the nature of the enactment, and the strictness or otherwise of the words in which it has been expressed. The courts should be less willing to extend expressed meanings if it is clear that the Act in question was designed to be restrictive or circumscribed in its operation rather than liberal or permissive. They will be much less willing to do so where the new subject matter is different in kind or dimension from that for which the legislation was passed. In any event there is one course which the courts cannot take under the law of this country: they cannot fill gaps; they cannot by asking the question, ‘What would Parliament have done in this current case—not being one in contemplation— if the facts had been before it?’, attempt themselves to supply the answer, if the answer is not to be found in the terms of the Act itself.”

The meaning of the terms “article” and “infringing copy”

565.

The parties dealt with these terms separately in their submissions, but there is a considerable overlap for reasons which will become clear. Each prays in aid the words of the statute, the statutory context, the decision in Sony Computer Entertainment Inc v Ball [2004] EWHC 1738 (Ch), [2005] FSR 9 (“Sony v Ball”) and the need for a purposive interpretation of the statute in support of their respective interpretations of the words “article” and “infringing copy”. Getty Images emphasise the “always speaking” principle, while Stability invites me to have regard to the record in Hansard of what was said in Parliamentary debates on sections 22 and 23 CDPA in determining the meaning of the term “article”.

566.

Getty Images began their submissions with the interpretation of the word “article” and so that is also where I shall begin.

The Interpretation of the word “article”

567.

There is no relevant statutory definition of “article”. Getty Images submit that the word “article” is a very general term with a broad meaning and that it does not have to denote a tangible thing. They point to a “news article” or an “article of the constitution”. They also point to the fact that in each of sections 22 and 23 CDPA the term “article” is used in conjunction with “an infringing copy”, i.e. “an article which is…an infringing copy”. This, say Getty Images, reinforces the wide meaning of “article”, since the word “copy” is sufficiently wide to cover both tangible and intangible things (as is borne out by sections 17(2) and 27(6) CDPA). Furthermore, they say that none of the activities which must be done with an article, as contemplated in sections 22 and 23 CDPA (e.g. importing, possessing, selling, distributing) are obviously aimed only at tangible things. These activities can also apply to an intangible article.

568.

Stability, on the other hand, points to Stroud’s Judicial Dictionary of Works and Phrases which identifies various statutes and authorities in which the term “article” appears. Stability says that the term “article” is not used in any of these examples in a sense that would include intangible matter or abstract information. Contrary to Getty Images’ submissions, Stability says that the words used to define types of secondary dealing in sections 22 and 23 CDPA in fact point towards the word “article” meaning only a tangible thing.

569.

It is clear from the extract from Stroud’s Judicial Dictionary relied on by Stability that, as one might expect, the word “article” will almost invariably take its colour from its surrounding context. On its own, it is not a word that has a clear meaning; it is no more than a generic term for an object, item or thing. It is commonly used in connection with tangible objects (such as an article of clothing or a household article) but that does not mean that it is incapable of applying to something intangible in the right context – for example a newspaper article or an article of faith. That the term is used in many statutes or contracts to refer to a moveable chattel or physical object of some kind does not appear to me to assist in connection with this particular statute.

570.

Sections 22 and 23 CDPA provide the colour. They apply to an “article” which is “an infringing copy of the work”. Such an “article” will be capable of being imported, possessed, sold, let for hire, offered or exposed for sale or hire, exhibited or distributed. However, it seems to me to be implicit in these sections that any article which falls within the definition of an “infringing copy” is intended for the purposes of this statute to be capable of being subject to these secondary types of dealing. Accordingly, to my mind, the first question for the court must be whether an infringing copy is capable of encompassing an intangible article. Arguments to the effect that an intangible article cannot be imported or possessed (although I shall return to them in a moment), do not really seem to me to assist in the abstract.

571.

It is common ground that sections 22 and 23 CDPA are to be construed in their statutory context. As Stability submits, this includes the context provided by Part 1 to the Act which is concerned with copyright. Section 27 CDPA provides the meaning of infringing copy. It makes clear that the words “infringing copy” in Part 1 are to be construed in accordance with “this section” and I have set out sections 27(1)-(3) above. The section is concerned with the circumstances in which an article will amount to an infringing copy (whether because its making constituted an infringement of copyright or because it has been imported into the United Kingdom and its making in the United Kingdom would have constituted an infringement). The section does not deal with what is meant by “copying” (or indeed “copies” or “copy”) – for that one must look to section 17 CDPA.

572.

Section 17 provides (insofar as relevant) as follows:

“17.— Infringement of copyright by copying.

(1)

The copying of the work is an act restricted by the copyright in every description of copyright work; and references in this Part to copying and copies shall be construed as follows.

(2)

Copying in relation to a literary, dramatic, musical or artistic work means reproducing the work in any material form. This includes storing the work in any medium by electronic means.”

(6)

Copying in relation to any description of work includes the making of copies which are transient or are incidental to some other use of the work” (emphasis added).

573.

As I said in my judgment on the Summary Judgment application, it is clear from section 17(2) that reproduction of the copyright work (for the purposes of primary infringement) can be “in any material form” and that it includes “storing the work in any medium by electronic means”. Thus Getty Images contend that owing to the use of Copyright Works in the training data (which had to be reproduced and stored locally or in the cloud), the making of Stable Diffusion involved reproducing the Copyright Works in a material form which (had it taken place in the United Kingdom) would have been contrary to section 17 CDPA 1988, because the unauthorised storage in electronic form of a copyright work will amount to an infringing reproduction of that work.

574.

It seems to me that when construing the meaning of “infringing copy” and also, therefore, “article” for the purposes of sections 22, 23 and 27 CDPA, one cannot ignore the context provided by section 17 as to what may be an infringing copy. The words of the statute must be construed having regard to their proper context so as to ensure coherence and consistency.

575.

As Getty Images submit, this approach is consistent with the approach adopted by Laddie J in Sony v Ball. In that case it was common ground that a Random Access Memory chip (or RAM chip) was capable of being an article, but the key question was whether a RAM chip that contained an infringing copy of digital data for only a fraction of a second could be an article (or as counsel put it, a “tangible substance”) given its ephemeral state. Laddie J held that it could. In setting out his reasoning, he referred to both sections 27 and 17 CDPA 1988 saying this:

“15.

…The silicon RAM chip is an article. When it contains the copy data, it is also an article. The fact that it did not contain the copy before and will not contain the copy later does not alter its physical characteristics while it does contain a copy. It is always an article but it is only an infringing article for a short time. There is nothing in the legislation which suggests that an object containing a copy of a copyright work, even if only ephemerally, is for that reason to be treated as not an article. On the contrary, the definition in s.27 points to the instant of making of the copy as crucial to the determination of whether or not it is an infringing article. An article becomes an infringing article because of the manner in which it is made. Whether it is an infringing article within the meaning of the legislation must be determined by reference to that moment. It matters not whether it is remains in that state, since retention as a copy is no part of the definition in the section

16.

This is also consistent with the provisions of s.17…

17.

In my view the meaning of the term “infringing copy” takes colour from its context in the Act. Just as a transient act of copying amounts to infringement, so an article which transiently contains a copy is an infringing copy for the purpose of this legislation. Taken together, these two subsections appear to suggest that even making a transient copy of a work can constitute making a reproduction “in a material form”. Thus RAM containing a copy of Sony’s copyright work is a reproduction in material form. It would produce an unwarranted inconsistency in the Act were that material form not to be considered an article for the purpose of s.27” (emphasis added).

576.

This final sentence is strongly supportive of the proposition that the concept of an “article” as used in the statute cannot be divorced from the concept of an infringing copy, or indeed from what can amount to an act of copying or a copy under the Act.

577.

Laddie J was not in that case directly concerned with (i) the question of what the word “article” means under the Act (there was no suggestion that a RAM chip could not be an article), or (ii) whether a copy stored on an intangible storage medium could still be an infringing copy (a RAM chip being a physical tangible means of storage). It does appear, however, that the argument proceeded on the assumption that the infringing copy must be tangible – hence the centrality of the question of whether a RAM chip containing a copy “is too short lived to be regarded as tangible”. Thus the judgment sheds some light on what amounts to an infringing copy when a work is stored by electronic means. Laddie J regarded the RAM chip as containing a copy of the work in “any material form” for the purposes of section 17(2). He found that the RAM chip (on which the work had been stored electronically, albeit for a short space of time), was an infringing copy, but it was only an infringing copy while the work was contained in it. Implicit in this judgment is an interpretation of infringing copy which (in the case of storage of an electronic work) includes (as Stability contends) the media on which it is stored.

578.

Must that media inevitably be tangible? Sony v Ball is not authority for such a proposition. Laddie J did not directly address that question. The words “in any material form” in section 17(2) are addressed at the reproduction of the copy, while the concept of storage in “any medium” could hardly be more broad or generic.

579.

At the time of the CDPA – as is common ground – storage media capable of storing electronic copies would have involved physical, tangible means of storage such as a floppy disk or magnetic tape (more recently a hard drive, flash drive or USB stick and, by the date of Sony v Ball, a RAM chip). However, in today’s world, electronic copies can be stored in an intangible “cloud” with no physical instantiation. Given the breadth of the words “any medium”, it seems to me that section 17(2) was plainly intended by Parliament to cover any means of storing electronic copies and thus must be capable of including modern means of storage such as cloud storage.

580.

I agree with Getty Images that the “always speaking principle” is of assistance in these circumstances. Stability does not suggest that the statute was intended to be “frozen” in time and I consider that modern storage methods in intangible media amount to a fresh set of facts which fall within the same genus of facts as those to which the original expressed policy has been formulated. The fresh set of facts arises by reason of the prevalence in the modern world of intangible electronic storage which has been brought about by enormous strides in technology since the date of commencement of the CDPA. The purpose of the Act – the protection of copyright owners – would, in my judgment, be fulfilled by an interpretation which encompassed modern technology.

581.

Stability’s submission that electronic copies of copyright works were known to Parliament in 1988 is true, but tilts at the wrong windmill. On Stability’s own submissions, intangible means of storage had not yet been invented, so the failure to make express provision for them in the Act cannot possibly be said to be a deliberate policy. I reject Stability’s suggestion that I am in any way assisted on this question by the provisions in the CDPA dealing with broadcasting and transmission, which have nothing whatever to do with the storage of electronic copies.

582.

I am fortified in this view by reference to the fact that, in a slightly different context, the CJEU has held that saving a copy of a work in storage space made available to a user in connection with cloud computing services constitutes a reproduction of that work (see Austro-Mechana Gesellschaft zur Wahrnehmung mechanisch-musikalischer Urheberrechte GmbH v Strato AG (C-433/20) EU:C:2022:217, [2022] ECDR 10, a case on Article 5(2)(b) of Directive 2001/29, at [15]-[18]).

583.

In the circumstances, an electronic copy stored in an intangible medium (such as the AWS Cloud) is, in my judgment, capable of being an infringing copy and thus also capable of being “an article”. I cannot see that Parliament could objectively have intended the use of the generic term “article” in sections 22, 23 and 27 CDPA to limit the scope for protection in relation to electronic copies (as provided for in section 17), including in circumstances where the means of storing such copies might change by reason of advances in technology.

584.

I must address a few of Stability’s additional arguments. While on the one hand accepting that the words of a statute must be seen in their proper context, Stability argues that the provisions of section 17 are of no assistance in construing the term “infringing copy”, because “an infringing copy” must be not just a copy but also an article under each of sections 22, 23 and 27 CDPA”. Stability submits that the requirement that the “copy” is an “article” is “an additional requirement which means that the secondary infringement provisions can only apply to copies in tangible form”. This is effectively just a different way of putting the point I have addressed above about the storage medium for electronic copies forming part of the infringing copy. It takes matters no further. In addition, to my mind this submission looks at the statute from the wrong end of the telescope. The relevant sections refer to “an article which is…an infringing copy” – this drafting neither suggests nor implies that in addition to being a copy, an infringing copy must also be something else – on Stability’s case, an article in tangible form. In my judgment this interpretation is neither consistent with the words of sections 22, 23 and 27, nor with a proper understanding of those words in their context (which includes section 17).

585.

I am also unpersuaded by Stability’s argument that the various acts of secondary infringement can only apply where there is a tangible article. While the concepts of “importation” and “possession” (on which Stability focused) may ordinarily connote physical transfer or control of a tangible object, I am of course only concerned with those concepts as they are used in the context of this statute. None of the cases to which I was referred by Stability assists on that point:

i)

As to “importation”, Stability relied on LA Gear Inc v Hi-Tec Sports Plc [1992] FSR 121, in which Morritt J observed at [30] that “the importation must have occurred when the goods were physically received in this country…”, but a reading of the judgment confirms that this observation is expressly tied to the facts of that case (which concerned the importation of shoes). Morritt J was not concerned with the scope of the concept of importation for the purposes of sections 22 and 23 CDPA.

ii)

As to “possession”, Stability relied upon authorities concerned with common law liens or claims in detinue and conversion for the propositions that intangible information cannot be “possessed” (see e.g. Your Response Limited v Datateam Business Media Limited [2014] EWCA Civ 281, [2015] FSR 23 (“Your Response”)) and that there is no property in information (see Thaler v Comptroller General of Patents, Trade Marks and Designs [2021] EWCA Civ 1374, [2021] RPC 19 per Arnold LJ at [125]). However, while Your Response certainly makes plain that (consistent with the decision of the House of Lords in OBG Ltd v Allan [2007] UKHL 21) “the common law draws a sharp distinction between tangible and intangible property” (per Moore-Bick LJ at [15]), the Court of Appeal in that case was not considering the meaning of “possession” in the context of the CDPA. Indeed, in a paragraph on which Stability partially relied (at [42]), Floyd LJ expressly acknowledged the scope for intangible information to give rise to intellectual property rights, such as database right and copyright, notwithstanding that more generally the law has been reluctant to treat information itself as property.

586.

Further, that an “article” which is an infringing copy can be “intangible” in that it can encompass electronic copies (which may not be “stored” in any tangible form), finds support, as Getty Images submits, in section 27(6) CDPA which provides that:

“In this Part “infringing copy” includes a copy falling to be treated as an infringing copy by virtue of any of the following provisions”

By the time of closing submissions, it was accepted by Getty Images that only one of the provisions listed under s. 27(6) was relevant as having been present in the statute at the time of its enactment (see R (Brown) v Secretary of State for the Home Department [2015] UKSC 8, [2015] 1 WLR 1060 per Lord Toulson JSC at [24]), namely:

“Section 56(2) (further copies, adaptations, &c. of work in electronic form retained on transfer of principal copy),…”

587.

Section 56 (insofar as relevant) provides as follows:

“56.— Transfers of copies of works in electronic form.

(1)

This section applies where a copy of a work in electronic form has been purchased on terms which, expressly or impliedly or by virtue of any rule of law, allow the purchaser to copy the work, or to adapt it or make copies of an adaptation, in connection with his use of it.

(2)

If there are no express terms—

(a)

prohibiting the transfer of the copy by the purchaser, imposing obligations which continue after a transfer, prohibiting the assignment of any licence or terminating any licence on a transfer, or

(b)

providing for the terms on which a transferee may do the things which the purchaser was permitted to do,

anything which the purchaser was allowed to do may also be done without infringement of copyright by a transferee; but any copy, adaptation or copy of an adaptation made by the purchaser which is not also transferred shall be treated as an infringing copy for all purposes after the transfer”.

588.

As Stability argues, section 56(2) is a deeming provision: it deals with the situation where a purchaser (who has obtained a copy of an electronic work which he can legitimately copy for his own purposes pursuant to the terms of purchase) transfers the copy to a third party in the absence of any contractual provisions determining the then status of any copy or adaptations of the copy in the hands of the transferee or the original purchaser. Specifically it deems that any copy or adaptation made by the purchaser which is not transferred to the third party “shall be treated as an infringing copy” .

589.

Accordingly, in my judgment, Stability is right that this provision is designed to address a scenario that might “fall through the cracks” in the sense that it would otherwise be unclear what rights the transferee might have in the copy of the work upon transfer and what the status of any copy of the work retained in the hands of the purchaser following transfer might be. However, I disagree with Stability that section 56(2) is supportive of the proposition that an electronic copy could not be an infringing copy absent this deeming provision. To my mind, the section is not designed to address the question of whether an electronic copy can be an infringing copy (that is already plain from section 17). All this section is doing is ensuring that, in the specific factual scenario postulated, an electronic copy shall be treated as an infringing copy, notwithstanding that the purchaser originally obtained the right to copy the work. Interestingly, Stability submits that any copy retained by the purchaser “would [in 1988] inevitably involve storage on a physical medium” – but that of course circles back into the question I have already addressed.

590.

Pausing there to take stock, on a proper construction of the statute, I consider that an article, which must be an infringing copy, is capable of being an electronic copy stored in intangible form. Standing back, I agree with Getty Images that if the word “article” were construed as only covering tangible articles, this would deprive authors of protection in circumstances where the copy is itself electronic and it is then dealt with electronically. Not only would that be inconsistent with the words of the statute, but it would also be inconsistent with the general scheme of copyright protection which is to reward authors for their creative efforts. As Lord Bingham said in Designers Guild Ltd v Russell Williams (Textiles) Ltd [2000] 1 WLR 2416 at 2418:

“The law of copyright rests on a very clear principle: that anyone who by his or her own skill and labour creates an original work of whatever character shall, for a limited period, enjoy an exclusive right to copy that work. No one else may for a season reap what the copyright owner has sown”.

As Getty Images submit, the importance of this principle is particularly acute in the modern world where huge numbers of works and copies are in electronic, intangible, form, owing to developments in technology. The “always speaking” principle of interpretation appears to me to support a construction that facilitates protection for copyright owners.

591.

I reject Stability’s contention that the relevant legislative provisions (insofar as they concern the interpretation of “article”) are “ambiguous, obscure or, on a conventional interpretation, lead to absurdity” such that pre-legislative statements made in Parliamentary debates are admissible on this question of statutory construction (see R (Coughlan) v Minister for the Cabinet Office [2022] UKSC 11, [2022] 1 WLR 2389 (“Coughlan”) per Lord Stephens JSC at [14] referring to the well-known conditions set out by Lord Browne-Wilkinson in Pepper v Hart [1993] AC 593 at 640). While the interpretation of the word “article” is not entirely straightforward and, certainly in this case, has led to considerable debate, I do not consider that a close analysis of the statute (the primary source for interpretation) and an application of the well-known principles of statutory construction to which I was referred, gives rise to any ambiguity or obscurity. Accordingly, there is no need for me to extend an already very lengthy judgment by considering the extracts from Hansard to which Stability drew my attention and I reject Stability’s submission that some form of “reverse engineering” in relation to Hansard (involving looking at Hansard to see whether it is inconsistent with Getty Images’ submissions before determining whether there is any ambiguity in the statute) would be appropriate. Even if I am wrong about that, I am inclined to agree with Getty Images that the Hansard materials on which Stability seeks to rely are not “clear and unequivocal on the point of interpretation which the court is considering” (Coughlan at [14]) owing to their apparent focus on broadcasts and cable programmes.

The Interpretation of “an infringing copy”

592.

That an “infringing copy” may be an electronic copy stored in an intangible medium, such that an “article” for the purposes of sections 22 and 23 CDPA encompasses an electronic, or intangible copy, is not however the end of the argument. Stability contends that whether an article is tangible or intangible, an infringing copy must be a copy; i.e. a reproduction of a Copyright Work. This is central to Stability’s case because it says that where, as is common ground, the model weights of the various Stable Diffusion versions do not store the visual information in the Copyright Works, they cannot amount to a copy.

593.

Getty Images say that Stability’s interpretation is obviously wrong, because the words of section 27(2) and (3) CDPA make plain that what is required if an article is to be an infringing copy is that “its making constituted an infringement”. They submit that there is no requirement that the article thus made must continue to retain a copy or copies of the work and they point to the fact that section 17(6) makes clear that “copying in relation to any description of work includes the making of copies which are transient or are incidental to some other use of the work”. They continue to emphasise the importance of a purposive construction. They also rely upon Sony v Ball at [15] (set out in full above) focussing specifically on Laddie J’s observations that:

“[a]n article becomes an infringing article because of the manner in which it is made. Whether it is an infringing article within the meaning of the legislation must be determined by reference to that moment. It matters not whether it remains in that state, since retention as a copy is no part of the definition in the section” (emphasis added).

594.

At the heart of Getty Images’ case is the submission that, unlike the facts of Sony v Ball, “[i]n this case the time of making of the copies of the Copyright Works coincides with the making of the article, i.e. the AI model. Thus unlike the [RAM] chip, as soon as it is made, the AI model is an “infringing copy”” (emphasis added).

595.

Sections 22 and 23 CDPA are not addressed to the act of reproduction of a copy, but to the downstream dealings in an article which is “an infringing copy”. Section 27 provides for the meaning of “infringing copy” which is to be construed “in accordance with this section”. This led Getty Images in opening to submit that the scope for looking at the statutory context (at least for the purposes of construing the phrase “infringing copy”) is limited. I disagree, as will already be clear. Stability points out (in my view, correctly) that “infringing copy” is a composite phrase and that the fact that it is defined as a phrase does not negate the meaning of the constituent words. It drew my attention to G4S Plc v G4S Trustees Ltd [2018] EWHC 1749 (Ch) in which Nugee J (as he then was) had to construe the term “pensionable service”, which was a defined phrase in the Pensions Act 1995. At [26] he said this:

“I am conscious that in section 124(1) of the Pensions Act 1995 the expression “pensionable service” is itself a defined expression, and that ultimately what I have to construe are the words of the definition, not the words “pensionable service” as such. But where a statute (or indeed a contract) uses a defined expression, the ordinary meaning of the word or words defined is part of the material which can be used to construe the definition. The defined expression is seldom an arbitrary label; it is usually chosen as a distillation of the meaning of a concept more precisely stated in the definition: see Birmingham City Council v Walker [2007] 2 AC 262, para 11, per Lord Hoffmann and compare Chartbrook Ltd v Persimmon Homes Ltd [2009] AC 1101, para 17, again per Lord Hoffmann.”

596.

I agree with Stability that Nugee J’s reasoning is of direct application here. As I have already indicated, the word “copy” is not arbitrary – the statute provides in section 17 (entitled “Infringement of copyright by copying”) that it shall have a particular meaning and it seems to me that that meaning must inform the proper interpretation of the composite phrase “infringing copy”. This is entirely consistent with the approach of Laddie J in Sony v Ball as set out above, who expressly considered (at [15]) that “the meaning of the term “infringing copy” takes colour from its context in the Act”. It is also consistent, as Stability points out, with the fact that the concept of copying and copies is a fundamental concept in an Act which is concerned with the provision of protection against the copying of copyright works.

597.

Thus, it seems to me to be clear that an infringing copy must be a copy, as Stability submits; the essence of the infringement is that there has been an infringement of copyright by the reproduction of the work (including by its storage in any medium by electronic means) in any material form. Consistent with the decision in Sony v Ball, I cannot see how an article can be an infringing copy if it has never consisted of/stored/contained a copy. In Sony v Ball the RAM chip (which was always an article) was only an infringing copy while it contained the copy but not otherwise: it was therefore, in the words of the Judge, “only an infringing article for a short time”, i.e. for the time that it contained the copy.

598.

As Sony v Ball also makes plain, an article becomes an infringing copy when the act of reproduction occurs. From that moment the article is an infringing copy – but it ceases to be an infringing copy once it no longer contains the copy. Thus it was no part of Stability’s case, (notwithstanding that Getty Images wrongly sought to characterise Stability’s case in this way), that for an article to be an infringing copy it must continue to retain a copy or copies of a Copyright Work. Stability accepts that transient copying can be sufficient to give rise to an infringing copy, but it says that an infringing copy must (at least at some point) contain a copy.

599.

In reality therefore, the dispute between the parties as it finally emerged in closing, really turns on whether an article whose making involves the use of infringing copies, but which never contains or stores those copies, is itself an infringing copy such that its making in the UK would have constituted an infringement. Taking the specific facts with which I am concerned, is an AI model which derives or results from a training process involving the exposure of model weights to infringing copies itself an infringing copy?

600.

In my judgment, it is not. It is not enough, as it seems to me, that (in Getty Images’ words) “the time of making of the copies of the Copyright Works coincides with the making of the Model” (emphasis added). While it is true that the model weights are altered during training by exposure to Copyright Works, by the end of that process the Model itself does not store any of those Copyright Works; the model weights are not themselves an infringing copy and they do not store an infringing copy. They are purely the product of the patterns and features which they have learnt over time during the training process. Getty Images’ central submission that “as soon as it is made, the AI model is an infringing copy” is, accordingly, in my judgment, entirely misconceived. Unlike the RAM chip in Sony v Ball which became an infringing copy for a short time, in its final iteration Stable Diffusion does not store or reproduce any Copyright Works and nor has it ever done so. The fact that its development involved the reproduction of Copyright Works (through storing the images locally and in cloud computing resources and then exposing the model weights to those images) is of no relevance. Furthermore, that there is no requirement that an article which is an infringing copy must continue to retain a copy does not assist Getty Images, because it is implicit in the word “continue” that at some point the article has in fact contained an infringing copy. The model weights for each version of Stable Diffusion in their final iteration have never contained or stored an infringing copy.

601.

I agree with Stability that the concept of an infringing copy cannot be interpreted in the abstract without reference to the fundamental nature of a copy, as revealed by section 17. Such an interpretation precludes the potential for an article which has never contained a copy to be capable of being “an infringing copy”. But, sections 27(2) and (3) require the article that is made to be a “copy”. They are not concerned with a process which (while it may involve acts of infringement) ultimately produces an article which is not itself an infringing copy. Section 27(6) provides no assistance to Getty Images in this regard. I agree with Stability that when applying the ‘making’ concept to an intangible article (Footnote: 10), the conception or creation of the model weights (i.e. the process of creation of information) cannot be the relevant act of making – that can only be the act of production of the resulting intangible record of that information, which derives from that process of creation. It is that intangible record which, on Getty Images’ case, enters the UK and that is the specific article whose making (had it taken place in the UK) must have constituted an infringement of copyright.

602.

Finally, on this issue I am not persuaded that Getty Images’ “purposive interpretation” assists. The scheme of the statute is to provide protection to the owners of copyright works against acts of infringement involving copying. Ms Lane highlighted the importance of being able to prevent people from circumventing copyright protection by carrying out infringing acts abroad and then importing the fruits of those acts into the United Kingdom. However, where an AI model itself is not an infringing copy, it cannot have been the intention of Parliament that it should fall within the meaning of infringing copy in section 27. The secondary acts to which Getty Images objects thus fall outside the policy and object of the Act.

The factual questions

603.

In light of my findings on the law, it is not strictly necessary for me to make any determinations on the facts. I have already found that Stable Diffusion is not capable of being an infringing copy and thus sections 22 and 23 CDPA are not applicable. However, in case I am wrong in my interpretation of the CDPA, I nevertheless set out my findings on the facts as briefly as possible in this section.

604.

On the assumption that, contrary to my findings above, Stable Diffusion is capable of being an infringing copy for the purposes of section 27(3), then:

i)

There is evidence that Copyright Works were used for training the Model and Stability has made limited concessions to the effect that “at least some” Copyright Works were used in training the Model and that images from the Getty Images Websites were used in training. In a Response to an RFI, Stability made clear that it would not seek to challenge at trial the proposition that sample works A1-A17 from the SOCI were used in training versions 1.4, v2.x, all versions of XL and v1.6.

ii)

I find that, as Stability acknowledged in closing, the model weights for Stable Diffusion v2.x, XL 0.9, XL Base 1.0 and XL Turbo have all been made available by Stability for download on Hugging Face. This involves making intangible copies of these Models available for download onto a local device in the UK. Given that I have found that an article may be intangible, I accept that downloads of the Model amounted to importation for the purposes of the Act and that they have taken place within the United Kingdom. Stable Diffusion is an open source model which is available for anyone to download. Mr Auerhahn agreed in cross-examination (and I find) that there will have been many downloads of Stable Diffusion v2.x and SD XL on to computers in the UK (perhaps running into millions). This was not disputed by Stability in closing submissions. Under section 27(3) the importation need not be by the defendant and so it matters not whether the download is characterised as an act of the defendant or otherwise.

iii)

Getty Images accept (and I find) that gaining access to the Model via DreamStudio does not involve importation or even transfer of a copy of Stable Diffusion to the UK because DreamStudio is a hosted service that is provided from servers located outside the UK. No copy of the Model is ever provided to the user and all inference and output synthesis takes place outside the UK. Whatever my decision on the interpretation of the statute, there could never be any act of secondary infringement by reason of the provision of remote software services. The case on secondary infringement under sections 22 and 23 CDPA must fail in relation to DreamStudio. Given the way in which the Developer Platform works, the same must equally apply to it. In closing there was some discussion about Stability’s case that, since 14 December 2023, SD XL Turbo has been made available for download on new licensing terms. This appears to have been part of an AI model licensing programme which (given the period it has been available) I find on balance has involved downloads in the UK.

iv)

As I understood Getty Images’ submissions on the facts in closing, it contends that it is “clear that, at the very least, Stability has distributed Stable Diffusion in the course of business via Hugging Face”. I agree in relation to v2.x, SD XL and v1.6. I accept that Stability is responsible for these acts of distribution which (if the other statutory requirements are met) thereby amount to acts of secondary infringement pursuant to section 23(d) CDPA. The same applies in relation to downloads from the Model licensing programme.

605.

As to the question of whether (if Stable Diffusion is an infringing copy) Stability had knowledge or reason to believe that it was, Stability’s Defence focuses on the fact that Stability did not know or believe that the Model reproduced any element of the training dataset or any of the images contained within it and that it did not otherwise know of any facts or circumstances which would have put a reasonable person on notice of the contrary. However, I can only address this issue on the assumption that Stable Diffusion is an infringing copy because of the training process (i.e. that, contrary to my finding above, the reproduction and use of infringing copies during training is sufficient to satisfy the statutory test under section 27(3)). On that basis, the point can be dealt with shortly by simply setting out what Stability did know or have reason to believe.

606.

Although Stability failed to call any evidence from a witness designed to address the issue of knowledge, in my judgment there is available evidence that Stability knew or had reason to believe:

i)

that works of the sort with which this case is concerned are protected by copyright and that acts of reproduction, communication to the public, importation and possession of copies in the course of business are prohibited without the consent of the copyright owner – this much was admitted by Stability in its Defence. Mr Konstantinov confirmed that “copyright was something that people were talking about generally in the AI world”, while Ms Bakshandaeva confirmed in her evidence that she thought she was aware that stock images were protected by copyright;

ii)

that Stable Diffusion was being trained on LAION datasets and that these had been produced by filtering the Common Crawl public web archive for images and storing them alongside their HTML alt-text (as confirmed by the Model Cards, by an internal Finance Department document dated 27 June 2023 and by the evidence of Professor Farid);

iii)

that the training data had been scraped from the internet without the consent of copyright owners; and

iv)

that the training data included works bearing the Getty Images watermark – had that not been the case, Stable Diffusion could not have produced output bearing the watermarks*. Mr Wagrez confirmed that he understood that it was “problematic having Getty Images in the dataset.”

607.

Furthermore, I accept Getty Images’ submissions that the latter two points are supported by disclosure documents from the trial bundles. In addition to evidence to which I have already referred in this judgment, I refer specifically to:

i)

an internal Stability Chat exchange taking place on 7 June 2022 in which the individuals involved in the chat (including Mr Mostaque) acknowledge the need to get “watermarks out of the dataset”, the potential (depending on “the PR/legal fight”) for it to be possible only to train on “public domain data”, the fact that “…people will go ‘this model was trained on our copyrighted data and we were not compensated for its use’ when they perceive it to threaten their livelihoods” and that “whether or not we turn out to be legal in the clear for using it in the end, there is going to be a ton of noise over it”.

ii)

An internal Stability Chat exchange taking place on 4 February 2023 in which the individuals involved in the Chat (including Mr Rombach) discuss options for “watermark removal” from the training set and refer to images with watermarks as making up “half the data”.

iii)

An email of 8 February 2023 from Joe Penna of Stability to Kenneth Lee and others at Stability, stating “At the moment we’re finding every model that Stability has available in staging & development (DeepFloyd, Stable Diffusion XL…) is not releasable due to issues with: Copyright (replicating logos, copyrighted artworks, etc etc etc…)”,

iv)

An internal Stability Chat exchange on 10 March 2023 in which Conner Ruhl of Stability refers to experiencing “watermarks and furniture…20% of the time”.

v)

An internal Stability Chat exchange from 6 April 2024 in which Mr Vencu explains that “urls with getty in their domain are clearly coming from a getty website”.

608.

Furthermore, a letter of claim was sent to Stability on 16 January 2023 asserting that Getty Images Content (said to be 12 million links) had been scraped from the internet and used to train AI models without consent which were now available to users via DreamStudio, thereby infringing copyright (amongst other things). The Particulars of Claim were first served on 12 May 2023.

609.

In closing, Ms Lane invited me to draw an inference in relation to the issue of knowledge by reason of Stability’s decision not to call Mr Vencu. While this is an issue in respect of which there might have been more scope for such an inference, I do not consider it would have been necessary, given the evidence to which I have referred above. Furthermore, I note that the Agreed Decision Tree for the Secondary Infringement Claim provided by the parties after the trial indicates that knowledge is admitted by Stability in the event that Getty Images’ interpretation of the CDPA were to be accepted. I had not understood that to be the position at trial, which is why I have made the necessary findings of fact.