Copyright and AI: Consultation

The following are my submissions to the UK Government’s Consultation on Copyright and Artificial Intelligence.

The background document which informs the questions and details the various options can be found at this page on the UK Government website.

Some of the questions were outside my area of expertise, and here I have replied with “Unsure” or similar.

Update: Just 24 hours after my submission to the consultation and there's talk of a partial climbdown by the government ;)

Questions 1-3 were personal information about me, so we start with question 4:

Copyright and Artificial Intelligence

Question 4. Do you agree that option 3 - a data mining exception which allows right holders to reserve their rights, supported by transparency measures - is most likely to meet the objectives set out above?

Answer: No, option 3 is unfairly balanced in favour of AI developers, and notably fails to address the issue of orphaned works (where the creative work being consumed by an AI developer is hosted on a third party site - possibly without the knowledge of the artist - with little or no identifying information that would allow the artist to opt out or an AI developer to know whether the artist of the piece had opted out.)

Question 5. Which option do you prefer and why?

Answer: Option 1: Strengthen copyright requiring licensing in all cases

Our proposed approach: Exception with rights reservation

Question 6. Do you support the introduction of an exception along the lines outlined in section C of the consultation?

Answer: No

Question 7. If so, what aspects do you consider to be the most important?

Answer: N/A - I do not support the introduction of an exception along the lines outlined in section C of the consultation

Question 8. If not, what other approach do you propose and how would that achieve the intended balance of objectives?

Answer: The law should be very clear that AI models could only be trained on copyright works in the UK if they have an explicit licence from the copyright holder to do so.

Question 9. What influence, positive or negative, would the introduction of an exception along these lines have on you or your organisation? Please provide quantitative information where possible.

Answer: As a UK artist it would have a large negative influence, because my artwork - especially if it has been uploaded to third party sites without my knowledge and without accreditation - could be copied into the AI model without due recompense from the big AI development corporations.

Question 10. What action should a developer take when a reservation has been applied to a copy of a work?

Answer: The developer should not include the copy of the work in their AI model in any way. If they felt it important to include it then they should contact the copyright holder of the work and seek direct permission, which might involve negotiating financial recompense.

Question 11. What should be the legal consequences if a reservation is ignored?

Answer: The AI model should be made unlawful to use until the work was expunged from the AI model. If it is impossible to expunge the work from the AI model - or if the copyright holder chooses this option - then the AI developer should be required to pay compensation to the copyright holder of the work at a level which is at the highest end of the spectrum of licensing costs, along with all reasonable costs for the time of the copyright holder in pursuing their claim.

Question 12. Do you agree that rights should be reserved in machine-readable formats? Where possible, please indicate what you anticipate the cost of introducing and/or complying with a rights reservation in machine-readable format would be.

Answer: This seems to be a sensible suggestion for sources such as a website controlled by the copyright holder, but fails to address the issue of orphaned works (where the creative work being consumed by an AI developer is hosted on a third party site - possibly without the knowledge of the artist - with little or no identifying information that would allow the artist to opt out or an AI developer to know whether the artist of the piece had opted out.)

If the machine-readable format was either similar to, or an extension of, the standard robots.txt format then introducing and maintaining this format would have a negligible cost. If the machine-readable format were a more complex SOAP or REST API then costs to both copyright holder and AI developer would significantly increase.

Technical Standards

Question 13. Is there a need for greater standardisation of rights reservation protocols?

Answer: The robots.txt standard is very widely adopted, and adherence to it (for site-wide level communication of "do not crawl this website") should be made mandatory for AI developers. However robots.txt currently needs to identify each AI crawler (user agent) in order to refuse it permission. An extension is required to be able to block or allow all AI crawlers in one go.

An extension of the widely adopted sitemap.xml standard - currently used for search engine crawling by the likes of Google - could be developed for a more page-by-page (URL by URL) set of rights reservation notices.

Question 14. How can compliance with standards be encouraged?

Answer: By making adherence to them mandatory for AI developers, and by providing "plug-ins" to add and maintain them for popular website frameworks (e.g. WordPress) available for free.

Question 15. Should the government have a role in ensuring this and, if so, what should that be?

Answer: If the AI developer has added works to their AI model where an easily maintainable rights reservation protocol (e.g. robots.txt) has forbidden them, then the AI model should be made unlawful to use until the work was expunged from the AI model. If it is impossible to expunge the work from the AI model - or if the copyright holder chooses this option - then the AI developer should be required to pay compensation to the copyright holder of the work at a level which is at the highest end of the spectrum of licensing costs, along with all reasonable costs for the time of the copyright holder in pursuing their claim.

Contracts and licensing

Question 16. Does current practice relating to the licensing of copyright works for AI training meet the needs of creators and performers?

Answer: No, as mentioned in section C.3 Technical standards standard terms and conditions for some services require works to be licensed on a broad basis which provide little choice to their users. It may be difficult for right holders to control the use of their work with AI, as terms relating to this may be bundled in with other contractual terms or framed in broad or vague language.

Question 17. Where possible, please indicate the revenue/cost that you or your organisation receives/pays per year for this licensing under current practice.

Answer: Not licensing any work at the moment.

Question 18. Should measures be introduced to support licensing good practice?

Answer: Yes, but I am not sure if these will be voluntarily taken up by the corporations seeking the licence.

Question 19. Should the government have a role in encouraging collective licensing and/or data aggregation services?

Answer: No

Question 20. If so, what role should it play?

Answer: N/A

Question 21. Are you aware of any individuals or bodies with specific licensing needs that should be taken into account?

Answer: It would be mainly individuals, for example individual artists who don't have legal departments

Transparency

Question 22. Do you agree that AI developers should disclose the sources of their training material?

Answer: Absolutely yes

Question 23. If so, what level of granularity is sufficient and necessary for AI firms when providing transparency over the inputs to generative models?

Answer: A basic request could be website-specific as to where the original creative work was copied from, but a copyright holder (artist, writer, etc) would need to be able make more specific queries if, for example, the website is a social media platform with millions of users.

Question 24. What transparency should be required in relation to web crawlers?

Answer: All AI web crawlers should have a clearly named User Agent String that allows them to be easily identified as an AI web crawler. They should also provide a URL of a page where more information could be discovered about what their crawler does (this is standard practice for “good agents”)

Question 25. What is a proportionate approach to ensuring appropriate transparency?

Answer: Lack of transparency will inevitably be used to hide copyright infringement, so the greatest possible transparency should be made mandatory.

Question 26. Where possible, please indicate what you anticipate the costs of introducing transparency measures on AI developers would be.

Answer: These would be low

Question 27. How can compliance with transparency requirements be encouraged, and does this require regulatory underpinning?

Answer: They would require regulatory underpinning in order to ensure compliance.

Question 28. What are your views on the EU’s approach to transparency?

Answer: I am unable to comment until development of the template to support reporting has been completed and released to the public.

Wider clarification of copyright law

Question 29. What steps can the government take to encourage AI developers to train their models in the UK and in accordance with UK law to ensure that the rights of right holders are respected?

Answer: Provide clear legal routes for AI developers to request permission from the copyright holder first and, if permission is agreed, compensate the copyright holder

Question 30. To what extent does the copyright status of AI models trained outside the UK require clarification to ensure fairness for AI developers and right holders?

Answer: If an AI model has not been trained in compliance with UK law then it should not be allowed to be used in the UK

Question 31. Does the temporary copies exception require clarification in relation to AI training?

Answer: Yes, definitely - the temporary copies exception is a loophole that could allow work to be copied and consumed by an AI model with the copyright holder’s permission. Its original purpose - e.g. copies held in browser caches or displayed on computer screens - should be maintained, but some clarification is needed to restrict it being used as a loophole by AI developers.

Question 32. If so, how could this be done in a way that does not undermine the intended purpose of this exception?

Answer: Not sure.

Encouraging research and innovation

Question 33. Does the existing data mining exception for non-commercial research remain fit for purpose?

Answer: Broadly yes

Question 34. Should copyright rules relating to AI consider factors such as the purpose of an AI model, or the size of an AI firm?

Answer: No, these might be taken into account by the copyright holder (artist, writer, etc) when negotiating a cost for allowing the AI firm to add the work to the AI model, but they should not be determined by law.

CGW Policy Option 0: No legal change, maintain the current provisions

Question 35. Are you in favour of maintaining current protection for computer-generated works? If yes, please explain whether and how you currently rely on this provision.

Answer: No, I am broadly in favour of Option 2: Remove specific protection for computer-generated works

Question 36. Do you have views on how the provision should be interpreted?

Answer: N/A

CGW Policy Option 1: Reform the current protection to clarify its scope

Question 37. Would CGW legislation benefit from greater legal clarity, for example to clarify the originality requirement? If so, how should it be clarified?

Answer: No, I am broadly in favour of Option 2: Remove specific protection for computer-generated works

Question 38. Should other changes be made to the scope of CGW protection?

Answer: N/A

Question 39. Would reforming the CGW provision have an impact on you or your organisation? If so, how? Please provide quantitative information where possible.

Answer: No impact

CGW Policy Option 2: Remove specific protection for CGWs

Question 40. Are you in favour of removing copyright protection for computer-generated works without a human author?

Answer: Yes. There is no need to enforce copyright protection for computer-generated works without a human author

Question 41. What would be the economic impact of doing this? Please provide quantitative information where possible.

Answer: N/A

Question 42. Would the removal of the current CGW provision affect you or your organisation? Please provide quantitative information where possible

Answer: Minor positive effect

Infringement and liability relating to AI-generated content

Question 43. Does the current approach to liability in AI-generated outputs allow effective enforcement of copyright?

Answer: Yes, but uncertain

Question 44. What steps should AI providers take to avoid copyright infringing outputs?

Answer: They should only train their AI models on works which have been explicitly licensed to them for that use by the rights holder

AI Output labelling

Question 45. Do you agree that generative AI outputs should be labelled as AI generated? If so, what is a proportionate approach, and is regulation required?

Answer: Yes - a small selection of “levels” could be employed from “made completely by AI” to “first draft made by AI but extensive editing by a human”, and regulation is definitely required.

Question 46. How can government support development of emerging tools and standards, reflecting the technical challenges associated with labelling tools?

Answer: provide clear guidelines, which should be enforced on large platforms (the social media platforms with millions of users). AI Companies should pay for these.

Question 47. What are your views on the EU's approach to AI output labelling?

Answer: I approve of the EU approach that requires providers of AI systems to mark their output as AI-generated content.

Digital replicas and other issues

Question 48. To what extent would the approach(es) outlined in the first part of this consultation, in relation to transparency and text and data mining, provide individuals with sufficient control over the use of their image and voice in AI outputs

Answer: Unsure

Question 49. Could you share your experience or evidence of AI and digital replicas to date?

Answer: N/A

Other emerging issues

Question 50. Is the legal framework that applies to AI products that interact with copyright works at the point of inference clear? If it is not, what could the government do to make it clearer?

Answer: No - where an AI product can interact with copyright works at inference the AI product should check (via a protocol such as the ones discussed earlier) whether it has permission from the copyright holder to interact with the copyright work

Question 51. What are the implications of the use of synthetic data to train AI models and how could this develop over time, and how should the government respond?

Answer: Unsure

Question 52. What other developments are driving emerging questions for the UK’s copyright framework, and how should the government respond to them?

Answer: Unsure