![]() |
[Home] [Databases] [World Law] [Multidatabase Search] [Help] [Feedback] | |
England and Wales High Court (Chancery Division) Decisions |
||
You are here: BAILII >> Databases >> England and Wales High Court (Chancery Division) Decisions >> Getty Images (US) Inc & Ors v Stability AI Ltd [2023] EWHC 3090 (Ch) (01 December 2023) URL: http://www.bailii.org/ew/cases/EWHC/Ch/2023/3090.html Cite as: [2023] EWHC 3090 (Ch) |
[New search] [Printable PDF version] [Help]
BUSINESS AND PROPERTY COURTS OF ENGLAND AND WALES
INTELLECTUAL PROPERTY LIST (ChD)
Fetter Lane London, EC4A 1NL |
||
B e f o r e :
Between :
____________________
GETTY IMAGES (US) INC GETTY IMAGES INTERNATIONAL UC GETTY IMAGES (UK) LIMITED GETTY IMAGES DEVCO UK LIMITED ISTOCKPHOTO LP THOMAS M BARWICK INC |
Claimants |
|
- and � |
||
STABILITY AI LTD |
Defendant |
____________________
Mr Nicholas Saunders KC and Mr Jaani Riordan (instructed by Bird & Bird LLP) for the Defendant
Hearing dates: 31 October & 1 November 2023
____________________
Crown Copyright ©
Mrs Justice Joanna Smith:
The Applications before the Court
a. reverse summary judgment and/or strike out in respect of various issues arising on the Claimants' claim ("the SJ Application"), namely
i. a claim of copyright and database right infringement arising from the Claimants' allegation that during the development and training of Stable Diffusion, Visual Assets and associated Captions, including the Claimants' Copyright Works, were downloaded on servers and/or computers in the United Kingdom ("the Training and Development Claim"). This application is pursued solely under CPR 24.2;
ii. a claim of secondary infringement of copyright said by the Claimants to arise by reason of the importation of an "article", namely the pre-trained Stable Diffusion software, into the UK ("the Secondary Infringement Claim"). This application is pursued under CPR 3.4(2)(a), alternatively CPR 24.2;
iii. trade mark infringement and passing off claims relating to the "output" of images generated by Stable Diffusion ("the Trade Mark and Passing Off Claims"), again pursued under CPR 3.4(2)(a), alternatively CPR 24.2. These claims are pleaded at paragraphs 56, 57, 58 and 61 of the Particulars of Claim.
b. further responses to be given to identified requests made in the Defendant's CPR Part 18 Request ("the RFI") dated 12 May 2023 ("the RFI Application"). The key focus of these requests is to obtain particulars of infringement not presently included in the Particulars of Claim as pleaded, which identifies a sample of Copyright Works for the purpose of subsistence and ownership only.
a. an order to fix the CMC, PTR, trial window and related directions ("the Fixing Application");
b. permission to amend the Particulars of Claim by the addition of a new claim relating to an image-to-image feature of Stable Diffusion which it is said permits users to make "essentially identical copies of copyright works" ("the Amendment Application"), now pleaded in paragraphs 33, 39, 50A and 50B, together with new Annex 8B, of the proposed Amended Particulars of Claim ("the Original Proposed Amendments").
The Evidence
The Relevant Principles
Summary Judgment
"i) The court must consider whether the claimant has a "realistic" as opposed to a "fanciful" prospect of success: Swain v Hillman [2001] 1 All ER 91;
ii) A "realistic" claim is one that carries some degree of conviction. This means a claim that is more than merely arguable: ED & F Man Liquid Products v Patel [2003] EWCA Civ 472 at [8];
iii) In reaching its conclusion the court must not conduct a "mini-trial": Swain v Hillman;
iv) This does not mean that the court must take at face value and without analysis everything that a claimant says in his statements before the court. In some cases it may be clear that there is no real substance in factual assertions made, particularly if contradicted by contemporaneous documents: ED & F Man Liquid Products v Patel at [10];
v) However, in reaching its conclusion the court must take into account not only the evidence actually placed before it on the application for summary judgment, but also the evidence that can reasonably be expected to be available at trial: Royal Brompton Hospital NHS Trust v Hammond (No 5) [2001] EWCA Civ 550;
vi) Although a case may turn out at trial not to be really complicated, it does not follow that it should be decided without the fuller investigation into the facts at trial than is possible or permissible on summary judgment. Thus the court should hesitate about making a final decision without a trial, even where there is no obvious conflict of fact at the time of the application, where reasonable grounds exist for believing that a fuller investigation into the facts of the case would add to or alter the evidence available to a trial judge and so affect the outcome of the case: Doncaster Pharmaceuticals Group Ltd v Bolton Pharmaceutical Co 100 Ltd [2007] FSR 63;
vii) On the other hand it is not uncommon for an application under Part 24 to give rise to a short point of law or construction and, if the court is satisfied that it has before it all the evidence necessary for the proper determination of the question and that the parties have had an adequate opportunity to address it in argument, it should grasp the nettle and decide it. The reason is quite simple: if the respondent's case is bad in law, he will in truth have no real prospect of succeeding on his claim or successfully defending the claim against him, as the case may be. Similarly, if the applicant's case is bad in law, the sooner that is determined, the better. If it is possible to show by evidence that although material in the form of documents or oral evidence that would put the documents in another light is not currently before the court, such material is likely to exist and can be expected to be available at trial, it would be wrong to give summary judgment because there would be a real, as opposed to a fanciful, prospect of success. However, it is not enough simply to argue that the case should be allowed to go to trial because something may turn up which would have a bearing on the question of construction: ICI Chemicals & Polymers Ltd v TTE Training Ltd [2007] EWCA Civ 725".
a. the decision of the Court of Appeal in Kawasaki Kisen Kaisha Ltd v James Kemball Ltd [2021] EWCA Civ 33, per Popplewell LJ at [18], a case concerned with the merits test to be applied on an application to serve a claim on a defendant out of the jurisdiction (that merits test being the same as applies in a summary judgment application), and Elite Property Holdings Ltd v Barclays Bank Plc [2019] EWCA Civ 204, per Asplin LJ at [41] and [42]. These passages appear to me to reformulate and restate the test as set out in Easyair. I note, in particular, however, that it is not enough to plead allegations which, if true, would establish a claim; there must be evidential material which establishes a sufficiently arguable case that the allegations are correct. This will include a sufficiently arguable case as to the drawing of an inference by reference to the facts.
b. ICI Chemicals & Polymers Ltd v TTE Training Ltd [2007] EWCA Civ 725, a case involving a short point on construction, in which Moore-Bick LJ made clear (at [13]) that "It is the responsibility of the respondent to an application of this kind to place before the court, in the form of a witness statement, whatever evidence he thinks necessary to support his case" (emphasis added). I pause to observe that the Training and Development Claim does not give rise to a short point on construction and nor is it a case in which I am being invited to "grasp the nettle" and decide a short point of law. In my judgment, one must be careful not to import propositions which are relevant to the situation in which the court is being asked to decide a discrete point of law, into situations where the court is being asked to grant summary judgment on the grounds that there is no evidential material which establishes a sufficiently arguable case and no real expectation of any such material being available at trial. In the latter case there is certainly no requirement that a respondent need place before the court all the factual evidence on which he wishes to rely at trial. As Lewison J said in Easyair, the court must take into account not only the evidence actually placed before it on the application for summary judgment, "but also the evidence that can reasonably be expected to be available at trial".
37. Indeed, the need to consider the evidence that might reasonably be expected to be available at trial was recently confirmed in HRH Emere Godwin Bebe Okpabi v Royal Dutch Shell plc [2021] UKSC 3, a case in which the Supreme Court had cause to consider the summary judgment test of real prospect of success. Lord Hamblen JSC pointed out (at [127]) that: "As Lord Briggs JSC stated in Vedanta [202] AC 1045, para 45: 'the court cannot ignore reasonable grounds which may be disclosed at the summary judgment stage for believing that a fuller investigation of the facts may add to or alter the evidence relevant to the issue'". Lord Hamblen formulated the question to be posed at [128] as follows: "�are there reasonable grounds for believing that disclosure may materially add to or alter the evidence relevant to whether the claim has a real prospect of success"
Strike Out
39. Once again, the principles are largely common ground between the parties. The court may strike out a statement of case pursuant to CPR 3.4(2)(a), read together with 3.4(1) which makes clear that a statement of case may include part of a statement of case, on the grounds that it "discloses no reasonable grounds for bringing the�claim". In considering a strike out application, the court must assume that the facts pleaded in the relevant statement of case are true and ask itself whether the claim advanced on the basis of those facts has a real prospect of success.
"22. As to the applicable test itself:(a) The court must consider whether the claimant has a "realistic" as opposed to a "fanciful" prospect of success: Swain v Hillman [2001] 1 All ER 91. A realistic claim is one that carries some degree of conviction: ED & F Man Liquid Products v Patel [2003] EWCA Civ 472. But that should not be carried too far: in essence, the court is determining whether or not the claim is "bound to fail": Altimo Holdings v Kyrgyz Mobil Tel Ltd [2012] 1WLR 1804 at [80] and [82].(b) The court must not conduct a mini-trial: Three Rivers District Council v Governor and Company of the Bank of England (No 3) [2003] 2 AC 1, in particular paragraph 95. Although the court should not automatically accept what the claimant says at face value, it will ordinarily do so unless its factual assertions are demonstrably unsupportable: ED & F Man Liquid Products v Patel; Okpabi and others v Royal Dutch Shell Plc and another [2021] UKSC 3, at paragraph 110. The court should also allow for the possibility that further facts may emerge on discovery or at trial: Royal Brompton Hospital NHS Trust v Hammond (No 5) [2001] EWCA Civ 550; Sutradhar v Natural Environmental Research Council [2006] 4 All ER 490 at [6]; and Okpabi at paragraphs 127-128.
23.. The other principle relevant to the present appeal is that it is not generally appropriate to strike out a claim on assumed facts in an area of developing jurisprudence. Decisions as to novel points of law should be based on actual findings of fact: see Farah v British Airways (The Times 26 January 2000, CA). In that case, the Court of Appeal referred back to the decision of the House of Lords in Barrett v Enfield DC [2001] 2 AC 550 where Lord Browne-Wilkinson said at 557e-g:
"In my speech in the Bedfordshire case [1995] 2 AC 633, 740 � 741 with which the other members of House agreed, I pointed out that unless it was possible to give a certain answer to the question whether the plaintiff's claim would succeed, the case was inappropriate for striking out. I further said that in an area of the law which was uncertain and developing (such as the circumstances in which a person can be held liable in negligence for the exercise of a statutory duty or power) it is not normally appropriate to strike out. In my judgment it is of great importance that such developments should be on the basis of actual facts found at trial not on hypothetical facts assumed (possibly wrongly) to be true for the purposes of the strike out"....
24.. The same point arose more recently in Vedanta Resources PLC & Another v Lungowe & Others [2019] UKSC 20. That was a case where the underlying duty of care was alleged against a parent company, rather than the company involved in the day�to�day running of the mine said to have caused the pollution. Lord Briggs said:
"48. It might be thought that an assertion that the claim against Vedanta raised a novel and controversial issue in the common law of negligence made it inherently unsuitable for summary determination. It is well settled that difficult issues of law of that kind are best resolved once all the facts have been ascertained at a trial, rather than upon the necessarily abbreviated and hypothetical basis of pleadings or assumed facts."
The Training and Development Claim
"The Defendant's development team includes a significant number of individuals who reside and work in the UK, who will be identified upon receipt of appropriate confidentiality undertakings, including the Defendant's lead generative AI developer. It is to be inferred that, during the development and training of Stable Diffusion, visual assets and associated captions, used for the development and training of Stable Diffusion, including Visual Assets (including Copyright Works) and Captions from the Getty Images Websites, were downloaded on servers and/or computers in the UK".
a. it is common ground that paragraph 43 is the only paragraph in the Particulars of Claim as it stands which seeks to connect the Training and Development Claim with any activity taking place within the United Kingdom. It does so, not by making any positive averment of fact, but rather by inviting an inference that during development and training of Stable Diffusion the Claimants' Copyright Works were downloaded on servers and/or computers in the UK.
b. on close analysis, the pleading provides no particulars of any separate or distinct process in relation to the "development" of Stable Diffusion. The paragraphs to which I have referred all focus on the "training" of Stable Diffusion by using the LAION-5B Dataset and the LAION Subsets, which is the essential means by which Stable Diffusion gained access to the Copyright Works. Nevertheless, paragraph 43 does appear to distinguish between development and training and, accordingly, I reject Mr Saunders' suggestion that the plea of "development" is "a dead duck". I note that Professor Farid's (for present purposes unchallenged) evidence supports the proposition that the development of diffusion models (in the broad sense) will include the development of the neural architecture of the model followed by several stages of iterative training and evaluation.
c. Mr Saunders invited me to focus on the date of release of the LAION-5B Dataset (referred to by Mr Rose in his first statement), namely 31 March 2022. It is his submission that given the pleaded case as to the training of Stable Diffusion using the LAION-5B Dataset, any claim pursuant to paragraph 43 can only be restricted to the time period between 31 March 2022 and 22 August 2022 (i.e. the date of launch of Stable Diffusion 1.0), or between 31 March 2022 and 24 November 2022 (i.e. the date of launch of Stable Diffusion 2.0). This he says is important when looking at the Location Issue for the purposes of the SJ Application. I reject this proposition. As Ms Lane correctly points out, the Claimants' pleaded case at paragraphs 34 and 40 expressly refers to training taking place "[f]rom a date unknown" and there is presently no separate temporal plea in relation to development. Furthermore, (i) Mr Rose's evidence very clearly refers to 31 March 2022 as the date on which LAION made the LAION-5B Dataset "publicly available", evidence which is supported by a screenshot from LAION's website presenting the new dataset; (ii) there is no evidence as to when LAION launched the LAION Subsets which it is specifically alleged were used to train Stable Diffusion; and (iii) the pleading expressly refers to the LAION-5B Dataset being created "with the support of the Defendant" � in other words, it is alleged that there was a close commercial relationship between LAION and the Defendant even before the public launch of LAION-5B. This appears to be borne out by the screenshot from LAION's website which thanks "our sponsors", including the Defendant, "for providing computing resources to produce this dataset".
"As regards the acts identified above at paragraphs 43 and 47, reproducing the whole or a substantial part of the Copyright Works or each or any of them, contrary to section 17 CDPA�".
"As regards the acts identified above at paragraphs 43 and 47, extracting a substantial part of the contents of the Database contrary to Article 7(1) of the Directive and Regulation 16(1) of the Regulations�".
a. Mr Mostaque's evidence is to the effect that there was no one at the Defendant working on image generation models in 2021, specifically that no one undertook training and development of Stable Diffusion in 2021. He explains that in November 2021, the Defendant secured access to a cluster of high performance computing resources from AWS based in datacentres in the United States. This is confirmed by Mr O'Donoghue in his first statement and by Mr Saltsman in his statement and appears to be uncontroversial. Mr Saltsman, who built the Defendant's AWS cluster in the United States, says that he used a very large number of high-performance graphical processor units ("GPUs") for that purpose.
b. Mr Mostaque explains that the Defendant's aim was to offer this "high-performance compute" to academic and non-profit researchers to promote the development and growth of open-source AI models; one of the first individuals to be given such access (in December 2021) was Katherine Crowson who was not an employee of the Defendant at this stage, but was working on training her own text-to-image model. Katherine Crowson subsequently became an employee of the Defendant in March 2022 but she, and another researcher hired at the same time, were living in the United States and were accessing Stability's AWS cluster for the purposes of "finishing Ms Crowson's model".
c. Mr Mostaque says that, towards the end of May 2022, he was contacted for the first time by Mr Rombach, lead author of a paper first published in December 2021 ("the Latent Diffusion Paper") on which Mr Mostaque says Stable Diffusion was ultimately based. Following his contact with Mr Rombach, Mr Mostaque says that he gave Mr Rombach and a colleague, Mr Esser (both of whom were living outside the UK), access to Stability's AWS cluster to develop the work they had done in connection with the Latent Diffusion Paper. He denies that they had any access to Stability's AWS cluster before that date. Mr Mostaque says that it was the work undertaken by Mr Rombach and Mr Esser using the Defendant's AWS cluster "to further develop their work on the Latent Diffusion Paper" that resulted in the release of Stable Diffusion in August 2022 � i.e. all training was done on the AWS cluster (a proposition that is consistent with an LMU Press Release of 1 September 2022). Mr Rombach was employed by the Defendant in September 2022 but continued to live outside the UK.
d. Mr Mostaque is based in the UK and says that he has never worked on the development or training of Stable Diffusion himself. His evidence is that it is his understanding that "none of [the Defendant's] current and/or former employees who have worked on the development and/or training of Stable Diffusion on behalf of [the Defendant] have been located in the UK". In his first statement he explains that he ruled out certain individuals simply based on their job titles and, in respect of the remainder, he states that (for individuals still employed by the Defendant) he asked his HR team to contact them directly to seek confirmation that they have never worked on Stable Diffusion, and (for individuals who are no longer employed by the Defendant) he asked his HR team to contact their manager to seek a similar confirmation. He gives no evidence as to any responses received, but says that "[a]fter completing that exercise, I am confident that no Stability employee based in the UK has ever worked on developing or training Stable Diffusion". In his second statement Mr Mostaque refers (as I have already said) to six additional individuals located in the UK, saying that he has asked his legal team to carry out the same process that the HR team conducted for the purposes of his first statement and that again, following that exercise he is "confident" that none of these individuals ever worked on developing or training Stable Diffusion.
e. Regarding the location of computing resources, Mr Mostaque says that the Defendant used its US-based AWS cluster for all of its work on Stable Diffusion. He accepts that employees who have accessed the Defendant's AWS cluster to work on Stable Diffusion will have done so from local machines, but he says that because none of these employees are or have been located in the UK, he can say with confidence that no local computing resources based in the UK have been used by the Defendant for its work on Stable Diffusion. Mr Mostaque says that he does "not see how training data could have been downloaded by Stability to local devices in the UK as part of its work on Stable Diffusion".
f. Mr O'Donoghue's evidence is said to be entirely supportive of the points made by Mr Mostaque. Mr O'Donoghue has reviewed financial and usage information made available to the Defendant via an online management portal provided by AWS, as provider of the Defendant's cloud services. This is referred to as the Stability AWS Account which includes a "Cost Explorer" function and a "Billing Dashboard". Subject to the caveat that such information does not differentiate between services provided in respect of Stable Diffusion and other services provided to the Defendant, Mr O'Donoghue's evidence is that it nevertheless confirms his understanding that the computing resources used by the Defendant for its training of Stable Diffusion "have at all times been located outside the United Kingdom". When purchasing services from AWS, the Defendant selects a region for those services from the available regions offered by AWS. The region selected will determine the physical location of the AWS data centres which provide the relevant services. Mr O'Donoghue says that the only AWS processing services that the Defendant has used for training Stable Diffusion have been in regions in the United States.
g. In his third statement, Mr O'Donoghue says that the Defendant has offices in Notting Hill, that employees use standard laptops (not high-powered workstations containing expensive GPUs) and in some instances shared keyboards and monitors. He confirms that the Defendant has no server racks containing GPUs, or the power supply, cabling, storage and cooling facilities this would require. He exhibits a list of IT assets purchased by the Defendant since 11 February 2022.
(i) Human Resources:
a. Screenshots from an interview with Mr Mostaque uploaded to YouTube on 13 August 2022 with the tag line "The Man behind Stable Diffusion", show Mr Mostaque responding to the question, "What is Stable Diffusion?" with the answer "this is an image model that we built off LAION-5B�the LAION guys were obviously here a while ago very close kind of working with us some of them are actually Stability employees as well" (emphasis added). In response to the question, "What if your model produces bad outputs?" Mr Mostaque replied "as an example we took some programmers out of Russia because they spoke out against the government there, you know and they came some came from the Ukraine as well and we fast tracked their residency in the UK" (emphasis added). Mr Saunders suggested that this evidence is not inconsistent with Mr Mostaque's existing evidence, that the Russian and Ukrainian individuals could have been working on other products with which the Defendant is involved and that it goes nowhere unless it is suggested that these individuals were involved in the development and training of Stable Diffusion. But that is being suggested and on one reading of Mr Mostaque's response to the interview questions, he was focussing specifically on individuals working on Stable Diffusion (as built from the LAION 5B Database). I note that the evidence in both of these documents is now deployed in the proposed amendment to paragraph 43 to support the inference that the Claimants seek to draw. Mr Mostaque has provided no explanation in relation to these examples and, indeed, it is worthy of note that none of his statements makes any reference whatever to LAION or to the involvement of LAION "guys" working together with the Defendant, or as its employees, in respect of the development or training of Stable Diffusion, or any other of the Defendant's products. Despite providing his second witness statement shortly before the hearing in circumstances to which I have already referred, Mr Mostaque does not address this evidence at all and does not deal with the obvious question raised as to whether the Defendant used independent contractors based ("here") in the United Kingdom in developing and/or training Stable Diffusion.
b. Screenshots from an interview with Mr Mostaque uploaded to YouTube by the TWIML AI Podcast on 12 December 2022 regarding Stable Diffusion 1.0 show Mr Mostaque saying "I literally took Ukrainian developers the houses were destroyed and brought them to the UK" (emphasis added). Once again this raises the possibility of developers working in the United Kingdom in connection with the development and training of Stable Diffusion, whether as independent contractors or otherwise. Again this evidence is not addressed in Mr Mostaque's second statement.
c. Furthermore, there is evidence to suggest that Mr Mostaque, who resides in London, may have been personally involved in developing Stable Diffusion, notwithstanding his evidence to the contrary. I note in particular the Defendant's (and Mr Mostaque's) statement to the court in what have become known for these purposes as the "Hodes Proceedings" (i.e. proceedings in the United States District Court Northern District of California San Francisco Division (case 3:23-cv-03481-MMC)), which describes Mr Mostaque as "a technology entrepreneur and one of the most recognised pioneers in generative artificial intelligence" and contends that from as early as May 2021 Mr Mostaque shared generative AI images with Mr Hodes (the plaintiff in the Hodes Proceedings) and posted publicly on Twitter (now known as "X") about the development of generative AI, including in October 2021, November 2021 and May 2022. Whether these posts relate to Stable Diffusion or not (and it is impossible to say), this evidence potentially implies a more hands-on approach by Mr Mostaque than he has admitted to in his statements in these proceedings. Mr Mostaque responds in his evidence that the work reflected in these tweets was undertaken in his personal capacity, again seeking to distance himself from any involvement with Stable Diffusion. However, the Hodes Proceedings refer to "Stability AI's generative imaging experiments" (not Mr Mostaque's generative imaging experiments). This all raises a question, as Ms Lane submits, over whether Mr Mostaque may himself have been working on Stable Diffusion from the UK, including in conjunction with "The LAION guys" � "working with us" (emphasis added).
(ii) Computer Resources
69. Professor Farid opines in his first report that the development of Stable Diffusion "may or may not have used AWS services and may have included using and evaluating data in the UK as well as moving data into the UK either on desktops or via other cloud services depending on where [the Defendant's] development team and their management was located". Professor Farid (quite properly) accepts that he cannot say where, or how, Stable Diffusion was developed but he observes that "it would have been possible to train the open-source version of Stable Diffusion on a desktop computer, and so it is feasible, and indeed likely (given the huge costs involved in using the powerful cloud services) that a combination of local laptop/desktop devices and AWS cloud services was used to develop the model�". Mr Saltsman rejects the proposition that the development of Stable Diffusion would have been more cost-effective if done on local machines and so rejects Professor Farid's evidence, but he does not say in terms that local machines were not used in the development of Stable Diffusion and he was not, in any event, employed by the Defendant prior to May 2023. Mr Mostaque's evidence is that the Defendant's employees who accessed the AWS cluster to work on Stable Diffusion will have done so from local machines although he says those individuals were all outside the UK. However, given the holes in his evidence as to individuals who might have worked on Stable Diffusion, there appears to me to be a prospect, which is more than merely fanciful, that local machines were used during this process in the UK.
a. Mr O'Donoghue identifies in his first statement that the AWS Billing Dashboard records a number of entries for data transfer services which have a description which contains reference to an AWS code associated with the Europe (London) region. The charge associated with these transfers is small (it is very slightly updated in his second statement) and confidential. Mr O'Donoghue has sought to investigate them with the assistance of a technical employee of the Defendant who has not provided a statement. He says that following discussions with this employee and with Mr Saltsman he is "confident" that none of the transfers concerned data which was used by Stability for its training of Stable Diffusion, but he says that he has "yet to determine exactly what they relate to".
b. Mr O'Donoghue's "confidence" that these transfers do not relate to data used by Stability for its training of Stable Diffusion is said by him to be justified because (i) none of Stability's current (or former employees) involved in training Stable Diffusion are (or were) based in the United Kingdom � this is not definitive in light of my analysis above; (ii) the Defendant's computing equipment located in the United Kingdom consists of standard laptops and PCs "without the high powered GPUs which are required for training large AI models such as Stable Diffusion" � this says nothing about development and nor does it address the evidence to which I have referred in paragraph 70 above. Finally, Mr O'Donoghue refers to confidential information, raising an issue about the relative size of the transfer. Professor Farid responds, making the point that the transfer relates to a not insignificant amount of data and that it could cover a large number of images (evidence which is not anywhere challenged by the Defendant). In his statement, Mr Saltsman explains that the transfers are between AWS data centres in the London region and three AWS data centres in the US; in other words, they are cloud to cloud transfers. Ms Lane submits that although this does not establish data being downloaded onto local computing devices, it does provide evidence of the transfer of data from the US into the UK and therefore potential evidence of copyright infringement by reason of a reproduction in the UK.
a. That there is a significant issue on the evidence over when work on Stable Diffusion commenced. Mr Mostaque says it began in late May 2022 in his evidence, but Professor Farid's (unchallenged) evidence is that, in his experience, a development period of 9 months "or more" to the final training stage is likely for software of this type. Potentially consistent with Professor Farid's views, but inconsistent with Mr Mostaque's evidence, is (i) the evidence in a tweet from Mr Mostaque dated 9 April 2023 in which he states that it took "18 months" to create "an org�that can make models of all media types, produce cutting edge infra code for the RLHF etc and run one of the largest public supercomputers in the world on 50-100x less spend than others is impossible. Except we did it & will keep scaling"; and (ii) the evidence in a screenshot of an interview with Mr Mostaque uploaded to YouTube on 6 January 2023 in which Mr Mostaque says that "it kicked off probably about 14 months ago saying let's do all the types of AI�from language models to�image and others". If the latter is a reference to work commencing on image models in about October 2021, then it also appears at odds with Mr Mostaque's statement that no one at the Defendant was working on image generation models in 2021. Whilst the tweet and interview may not be referring to Stable Diffusion, it is simply impossible at this juncture to say one way or the other; if they are referring to Stable Diffusion at the early stages of its development, then Mr Mostaque's evidence in his statement is plainly inconsistent.
b. There is also an issue on Mr Mostaque's evidence arising from the interview to which I have just referred, which suggests that he was first contacted by Mr Rombach in February 2022 (rather than at the end of May 2022 as he says in his statement). This apparent inconsistency is not dealt with in Mr Mostaque's second statement (and would not have been addressed by Mr Rombach's statement either).
c. I have already referred to the question-mark on the evidence over whether Mr Mostaque carried out work in a personal capacity or as a representative of the Defendant.
d. Mr Saunders says, essentially, that these factual questions are merely peripheral issues, of no relevance to the central (and only) question of whether Stable Diffusion was developed or trained by individuals in the UK and that they should not carry any real weight in the face of two statements verified by statements of truth from Mr Mostaque. I disagree. To my mind they are issues which raise obvious question-marks over the credibility of Mr Mostaque's evidence � evidence which it is essential the court should accept at face value if the Defendant has any hope of success on the SJ Application.
e. Furthermore, the question of when work commenced on the development and training of Stable Diffusion feeds directly into the question of whether the Defendant's evidence to the effect that there were no interventions from individuals working in the UK is complete, in the sense that it covers the whole of the relevant time period. Without knowing exactly what time period the court is dealing with (and what exactly the work involved � a point I have made earlier), it is impossible to determine that the Defendant's evidence is complete or entirely accurate. Equally, the apparent inconsistencies between Mr Mostaque's evidence and other available evidence casts doubt over the credibility of his statements generally. At this stage, it is impossible for the court to determine whether a witness will come up to proof at trial, and nor should it. Disclosure may establish that Mr Mostaque's evidence is unimpeachable. However, the court cannot ignore evidence available on a summary judgment application which might suggest that a statement of truth (confirming the truth of evidence which is central to the relevant factual issue) will be tested by cross examination at trial and may be found wanting.
f. A rather separate issue arises in respect of confidential evidence which suggests, by reference to an AWS Migration Plan, that computing resources other than AWS cloud resources were being used by the Defendant prior to their migration to AWS. There appears to be no explanation from the Defendant of this evidence, but if it shows computing resources other than AWS cloud resources being used for the early stages of development or training of Stable Diffusion prior to migration to AWS cloud resources, then that would appear to be inconsistent with the Defendant's evidence to date and certainly requires investigation. Looking closely at paragraphs 7 and 8 of Mr Saltsman's evidence, I reject the suggestion by Mr Saunders that it was intended to, or does, address this point. In reply, Mr Saunders said on instructions that this migration was, in reality, all US based because the particular processors involved are US based; however, he accepted that there is no evidence from the Defendant to this effect. On a summary judgment application of this sort � providing (on instructions) evidence of potentially relevant facts is simply not satisfactory. It is another indicator that the Location Issue requires a much fuller investigation than can be undertaken on such an application.
a. the precise nature of the development and training of Stable Diffusion, including when it commenced and where it was carried out. Given the various inconsistencies and queries on the evidence that I have already identified, I would expect this evidence to provide more detail around both the Defendant's human resources working in the UK and its computing resources;
b. the nature and purpose of the AWS transfers to London to which I have referred in paragraph 72 above;
c. the computer resources available to the Defendant prior to February 2022;
d. the nature and purpose of the migration that took place pursuant to the AWS Migration Plan;
e. the arrangements between the Defendant and LAION for the use of LAION employees, whether employed by the Defendant or as independent contractors, the computing assets that they used and the location from which they worked;
f. the arrangements between the Defendant and the "Russian" and "Ukrainian" programmers, their location and the computing assets that they used;
g. the precise nature of the work done by each and every one of the Defendant's UK development team, including any interventions in respect of the training or development of Stable Diffusion.
The Secondary Infringement Claim
"52.4 As regards the acts identified above at paragraphs 44 to 46, importing into the United Kingdom, otherwise than for private and domestic use, an article, namely Stable Diffusion, which is and which the Defendant knows or has reason to believe is, an infringing copy of the Copyright Works or each or any of them, contrary to section 22 of the CDPA; and/or
"22. Secondary infringement: importing infringing copyThe copyright in a work is infringed by a person who, without 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.
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.
�
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.�"
"An electronic database consists of structured information. Although information may give rise to intellectual property rights, such as database right and copyright, the law has been reluctant to treat information itself as property. When information is created and recorded there are sharp distinctions between the information itself, the physical medium on which the information is recorded and the rights to which the information gives rise. Whilst the physical medium and the rights are treated as property, the information itself has never been."
and by Arnold LJ in Thaler v Comptroller General of Patents, Trade Marks and Designs [2021] EWCA Civ 1374 at [125]:
"There are good reasons of principle and policy for this rule: it would have alarming consequences if there was property in information".
Sony v Ball
a. Contrary to the Defendant's submissions, I do not consider that Sony v Ball provides the answer. In brief summary only, it was common ground in that case that a silicon chip was capable of being an article. However, a key question for the court was whether a chip that contained an infringing copy of digital data for only a small fraction of a second could be an article given its "ephemeral" state. It was argued that the chip could not be regarded as a "tangible substance" in such circumstances. Laddie LJ did not grapple with the question of tangibility. Rather, he held (at [15]) that there is nothing in the legislation to suggest that an object containing a copy of a copyright work, even if only ephemerally, is for that reason to be treated as not being an article. He went on to say "[o]n the contrary, the definition in s.27 points to the instant of making 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."
b. Where the judge simply did not address the argument as to whether the chip (or indeed the information it held) could be regarded as 'tangible', I do not agree with the Defendant that it is necessarily implicit in this reasoning that the information stored on the chip could not be an article, or indeed that an article is limited purely to "tangible" objects. The court only addressed the issue of whether the silicon chip was an article in circumstances where the life of the copy on the chip was of extremely limited duration. I reject the suggestion that it is possible to extrapolate from this decision the conclusion that an article must inevitably be a tangible thing.
Wheat v Google
c. True it is that in Wheat v Google, the Chief Master expressed the view (at [48]) that "the acts referred to in section 23 CDPA are tangible acts of physical exhibition or distribution, not electronic transmission of information", a view which appears supportive of the Defendant's submission that an article can only be a physical tangible thing. However, as Ms Lane correctly points out, the hearing before the Chief Master concerned applications by the claimant (acting in person) to serve Google out of the jurisdiction. The Chief Master does not appear to have heard any detailed argument on this point � certainly he does not set any out � and he provides no reasoning in support of the view he has taken. He does not appear to have been invited to consider any authority or to engage in a textual or contextual analysis of the CDPA. Furthermore, he appears to have been concerned only with acts of "exhibition" and "distribution" and not with any of the additional acts of secondary infringement alleged in this case. In the circumstances, and with the greatest of respect to the Chief Master who appears to have received no assistance whatever on the point, I cannot attach any serious weight to his decision, which is not, in any event, binding on me.
The full statutory context
d. I consider there to be two points about the statutory context which may potentially be relevant but in respect of which I have not heard full argument at the hearing and/or cannot determine that the Claimants are bound to fail.
e. The first concerns the potential relationship between sections 22 and 23 and section 17 of the CDPA. 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, including "storing the work in any medium by electronic means". Ms Lane points out that there is no dispute for the purposes of the SJ Application that the making of Stable Diffusion involved reproducing the Copyright Works in a material form contrary to section 17 of the CDPA and she submits that this is of significance in the context of construing the meaning of "infringing copy" in section 27 and "article" in sections 22, 23 and 27. At first blush, I tend to agree, not least because in Sony v Ball, Laddie LJ appears to have taken the approach (at [16]-[17]) that when identifying an infringing copy for the purposes of section 27, it is necessary to have regard to what may be an infringing copy for the purposes of section 17. In particular, he observed that "[i]t would produce an unwarranted inconsistency in the [CDPA] were that material form [i.e., the chip containing the reproduction of the copyright work] not to be considered an article for the purpose of s.27".
f. Ms Lane referred me to a post Brexit EU authority to the effect that storage in a medium by electronic means can include cloud storage (see C-433/20, EU:C:2022:217 Austro-Mechana Gesellschaft zur Wahrnehmung mechanisch-musikalischer Urheberrechte GmbH v Strato AG [2022] ECDR 10 at [15]-[18]), submitting (by reference to Warner Music UK Ltd v TuneIn Inc [2021] EWCA Civ 441 at [90]-[91] per Arnold LJ) that this judgment is highly persuasive, albeit not binding.
g. Mr Saunders initially suggested in response to a question from me, that a "medium" (as identified in section 17) is similar to an "article", albeit that he did not appear to maintain that point. He acknowledged that he had not focussed on section 17 in connection with the interpretation of sections 22 and 23 of the CDPA and indicated that he might wish to come back to the point. However, despite Ms Lane dealing with the point in some detail, Mr Saunders did not revisit it, and (although the Defendant sought to distinguish between primary infringement in section 17 and secondary infringement in sections 22 and 23 of the CDPA in its skeleton argument) I have no clear explanation from the Defendant as to why section 17 is not an important and potentially relevant part of the contextual analysis. Albeit in a slightly different context, Mr Saunders showed me a passage from the decision of the Competition Appeal Tribunal in Umbrella Interchange Fee Claimants v Umbrella Interchange Fee Defendants [2023] CAT 49, in which the Tribunal decided (for detailed reasons) not to follow existing EU law. Mr Saunders acknowledged that he could not possibly invite me at this hearing (in so far as it may be relevant) to determine whether this court should follow post Brexit EU authority.
h. The second arises by reason of a point made by Ms Lane in her oral submissions but which did not appear in the Claimants' skeleton argument or its supplemental skeleton. Ms Lane took me in some detail to section 27(6) of the CDPA which provides that an "infringing copy" will include "a copy falling to be treated as an infringing copy" by virtue of a variety of other sections of the CDPA (introduced by way of amendment), there identified. She submits, correctly, that an essential part of the exercise of statutory construction is the consideration of the relevant words or phrases in their statutory context. She points to various provisions in these other sections which she submits (i) cast light on the meaning of "infringing copy" and, in particular, suggest that an "infringing copy" can include an "electronic" copy and, thus, an intangible copy; and (ii) suggest that not all of the secondary acts of infringement in section 23 are acts which must necessarily be associated with a tangible article � e.g. the act of selling or letting for hire.
i. In his reply, Mr Saunders (who had not previously been alerted to this point and was obliged to deal with it "on his feet") dealt with it only briefly. In essence he submitted that the Statutory Instruments that introduced the amendments to the primary legislation cannot have been intended to affect the true construction of existing provisions. However, he was unable to say whether there is any authority about the extent to which one can refer to later amendments to a statute to construe a pre-existing provision. This caused me at the time to observe that the matter did not appear to be straightforward and to express concern that the court did not have the material it needed to decide the issue. Mr Saunders responded that his primary position was that the argument had been raised too late and that he had not had an opportunity to look at it.
j. Be that as it may, it appears to me that it would be unsafe finally to decide an important point of statutory construction in such circumstances. Whatever the rights and wrongs of the Claimants raising this argument at the eleventh hour, it has been raised and the Defendant was not in a position fully to deal with it at the hearing. Mr Saunders continued to suggest that the sections to which I was taken by Ms Lane cannot possibly be relevant to the interpretation of the words "infringing copies" or "article", but I am not satisfied on the limited argument that I have heard that the Claimants have no real prospect of arguing successfully to the contrary at trial.
The authorities as to information not being property
k. I need say little about these beyond agreeing with Ms Lane that the authorities on which the Defendant relies are not directly relevant to the issue of construction that I am asked to determine. This was not disputed by Mr Saunders.
The Claimants' Amendment Application
"50A. From a date unknown but since at least in or around March 2023 the Defendant introduced an image-to-image feature which enables Stable Diffusion to generate a synthetic image output in response to an image uploaded by a user (either via Dream Studio or via a copy of Stable Diffusion downloaded from GitHub or via commercial offerings using Stable Diffusion), either with or without a text command. The user is able to determine how closely the synthetic image output matches the image prompt by the use of an "image strength" slider, with the maximum image strength and/or values approaching the maximum image strength providing images which comprise the whole or a substantial part of the image prompt.50B. Accordingly, when a user provides an image prompt (either via Dream Studio or via a copy of Stable Diffusion downloaded from GitHub or via commercial offerings using Stable Diffusion) comprising one of the Copyright Works or the Visual Assets and uses the maximum image strength or a value approaching the maximum image strength (the precise value to be identified by way of disclosure and/or evidence), the synthetic image output comprises the whole or a substantial part of the Copyright Work or Visual Asset. Pending the provision of disclosure and/or evidence, the Claimants rely upon the following facts and matters:
50B.1. In the case of Stable Diffusion downloaded from GitHub, the image strength is a value between 0.0 and 1.0 that controls the amount of noise that is added to the image prompt. As explained in the guidance on the GitHub website (a copy of which is attached hereto at Annex 8A) values that approach 1.0 allow for lots of variations in the synthetic image output by removing all pixel-based information. By contrast, values that approach 0.0 (i.e. the maximum image strength) result in a synthetic image output that reproduces the whole or a substantial part of the image prompt.
50B.2. When an image is uploaded by a user, it is first converted into noise, and this noise pattern is used as an input to the image synthesis process. Thus, the image-to-image feature uses both Stable Diffusion and the pixels of the user-supplied, uploaded image.
50B.3. Copies of synthetic images that were generated in response to image prompts comprising a Copyright Work or Visual Asset, together with the image strength indicated, uploaded via a copy of Stable Diffusion downloaded from GitHub are attached hereto at Annex 8B".
The relevant principles
Decision
a. as for the issue of communication to the public, Ms Lane correctly points out that Mr Saunders only raised concerns over this pleading for the first time on his feet, that there is a substantial amount of complex case law on communication to the public (none of which had been relied upon by the Defendant until Mr Saunders sought to rely in his oral reply upon an authority that had been included in the bundle for a different purpose) and that it has never previously been suggested by the Defendant that the Claimants cannot establish a real prospect of success in relation to communication to the public. No such case is advanced in respect of the allegations of primary infringement involving the text-to-image prompt, notwithstanding that exactly the same point (if correct) could presumably have been made on the SJ Application. In all the circumstances, I cannot conclude that the Claimants have no real prospect of success on this point. If this was a serious argument, I would have expected it to be identified well in advance of the hearing so that both sides could prepare to take the court to the relevant case law. As it is, the court has little more than a bare assertion from the Defendant that communication to the public cannot include the users with which this claim is concerned, notwithstanding its apparently inconsistent stance in not seeking to challenge the issue of communication to the public in connection with the text to image claim in its SJ Application.
b. As for the issue of authorising reproduction by users, I did not understand Mr Saunders to suggest that this plea in itself failed to give rise to an arguable case.
a. the Defendant dropped its complaint that no particulars have been given as to the sample images in Annex 8B in light of the fact that particulars have now been provided in the second statement of Mr Rose. I accept Mr Saunders' submission, however, that those particulars should now be reflected in the amended pleading.
b. the Defendant accepted that its criticism of the wording used in 50B.3, in particular the use of the word "uploaded" to refer to the entry of an image into Stable Diffusion, could be sensibly addressed by replacing the word "uploaded" with the word "entered". Once again, this should be reflected in the amended pleading.
c. there is no basis to refuse the application to amend on the grounds that the phrase "commercial offerings" is unclear, as submitted by the Defendant in its skeleton argument. This was a phrase used in paragraph 46 of the original pleading; no CPR Part 18 request was made in respect of it and it has now been explained in court as the provision of Stable Diffusion to third parties who are free to use it in any way they wish. Once again, I consider that this should be made plain on the face of the amended pleading. I should add that although Mr Saunders suggested for the first time in oral submissions that the reference in paragraph 50A to making Stable Diffusion available "via commercial offerings" could not possibly have a real prospect of success, this point was barely developed and was not raised in connection with the existing plea at paragraph 46 of the Particulars of Claim. In particular, he did not explain how or why the Claimants have no real prospect of establishing their plea of authorising reproduction by users in connection with this form of distribution. I am not going to refuse permission to amend on the basis of points raised at the eleventh hour in submissions which were not properly developed and in respect of which the Claimants had no proper opportunity to respond. That is particularly so given that, if the point has merit, it could also have been raised on the SJ Application, but was not.
What Order should be made on the Applications?
Footnotes:
Note 1 The Sixth Claimant is included as a representative party of the parties that have concurrent rights of action with the First Claimant under ss 101 and 102 CDPA 1988, pursuant to CPR 19.6. The Defendant reserves its position as to the way in which this representative claim is put, but it was not a live issue at the hearing. [Back]
Note 2 After the hearing there was a dispute between the parties over the terms of the order reflecting this agreement. The Claimants were concerned that the Defendant was seeking in the draft order to preserve its position to argue later that it would be entitled to disclosure relating to the exercise undertaken by the Claimants to produce the example synthetic outputs for the purposes of responding to the RFIs. This would not have reflected the concession the Defendant made at the hearing and I do not consider that it is now in a position to try to resurrect any such entitlement. However, my understanding of the Defendant�s position in light of further short written submissions is that it takes the view that any decisions concerning experiments and disclosure going beyond the process required for the purposes of responding to the RFI Application will be matters for the CMC at a later stage. This does not appear to me to cut across its concession made at the hearing. The final Order dated 1 November 2023 but sealed on 23 November 2023 reflects these points. [Back]
Note 3 A second dispute arose after the hearing as to the scope of an Amended Annex 8 to be provided by the Claimants which will include examples of the synthetic output relied upon using a prompt. At the hearing, the Claimants indicated that they would provide examples in relation to text prompts which �are not exactly the Getty Images caption�. However, the Defendant then stated that if the Claimants are seeking to maintain a case that the use of different words nevertheless prompts infringing output �it would be helpful to have an example of that�. I understood this to be accepted by Ms Lane, albeit that the transcript does not record her physical acknowledgment. Accordingly, the Order reflects the requirement to provide examples using a prompt which is not the whole of a Getty Images Caption, but also adds the requirement (in so far as the Claimants seek to maintain a case that words other than words used in the Getty Images Caption prompt the production of infringing output) to provide an example of the same. [Back]
Note 4 A third issue arose between the parties on the draft Order following the hearing as to the scope of the Claimants� agreement in this regard. I understood it to be the case that the Claimants were confirming at the hearing that they did not advance a positive case arising by reason of the decision inLouboutin v Amazon. I did not understand them to be accepting that they could never again refer to that case in the context of, for example, responding to any pleading from the Defendant. I amended the draft Order to reflect this position. Given the extent to which the parties have disagreed on numerous issues in these proceedings, I do not consider that this is an issue that should be left to be addressed, if necessary, by reference merely to the transcript of the hearing [Back]
5. Contrastive Language-Image Pre-training
6. I understand this case to have been heard on appeal in the Supreme Court, albeit not on this point.
7. Warner Music UK Ltd v Tunein Inc [2021] RPC 11 per Arnold LJ at [70]. The summary of CJEU case law in this paragraph runs to 15 sub-paragraphs. What appears from the summary is that 'communication to the public' must be interpreted broadly and, further, that determining whether there has been a communication to the public requires an individualised assessment based on various different factors, including the factual circumstances in which the works have been made available to recipients together with the number of persons who have had access to the works. Assuming these to be relevant factors under domestic law, they will be matters for evidence at trial; I cannot see how I can be satisfied by reference purely to this authority that there is no real prospect of the Claimants succeeding on the image-to-image claim. Further and in any event, I note that paragraph [70] is not part of the ratio of this decision � neither Rose LJ, nor Sir Geoffrey Vos MR agreed with it at paragraphs [181] and [191] respectively and neither of them considered it helpful to provide such a summary where it was unlikely to be valid for long and where the restatement of CJEU principles in domestic judgments were not strictly on point. To my mind this merely serves to emphasise that this court is in no position to determine that the Claimants' case of communication to the public has no prospect of success.