The ALPSP Copyright Committee has responded to the UK IPO AI TDM Consultation.

Section A

ALPSP is the international trade association which supports and represents organisations that publish scholarly and professional content. It has nearly 300 member organisations across 30 countries. Its vision is to foster a culture that will strengthen and advance academic and scholarly communications in all its forms. It does this by:

• Leading and enabling knowledge-sharing across its membership and the wider community
• Fostering the creation of international connections through networking and collaboration
• Informing and influencing policy and decision-making, advocating for society, not-for-profit and community-led publishing
• Promoting inclusivity and diversity, removing barriers to participation and increasing transparency.

This response will focus on the potential effects of a change to the laws on CGWs and TDM on the publishing industry generally and on ALPSP members in particular. We have responded to the questions included in the Consultation Response Form which affect the publishing industry directly. If it would be helpful, the ALPSP Copyright Committee would be happy to discuss our responses in more detail at any time.

Copyright – computer generated works (CGW)

1. Do you currently rely on the computer-generated works provision? If so, please provide details of the types of works, the value of any rights you license and how the provision benefits your business. What approach do you take in territories that do not offer copyright protection for computer-generated works?

Computer generated works are novel to the academic and scientific publishing industry; the first ‘AI-created’ book being published by Springer Nature in 2019. Although, so far, the industry has not had a great deal of experience in disseminating computer-generated works, there is increasing interest in using AI in the creation of works, and particularly in areas such as translation or indexing. This is currently a nascent field and, as such, our members believe it should be left to develop rather than make any legislative intervention at this point.

2. Please rank these options in order of preference (most to least preferred) and explain why.

Option Preference: 0

We believe that the current period of protection should be maintained. There is no positive reason to remove or reduce the current period of protection. As this is a very new field, we need to better understand how it develops, in order to understand what legislative intervention may be appropriate, particularly as it is not presently clear which types of work may be created in the future and fall under this provision. We have concerns that intervention at this stage may result in unforeseen or unintended consequences, including potentially stifling the market altogether.

3. If we introduce a related right for computer-generated works, as per option 2, what scope and term of protection do you think it should have? Please explain how you think this scope and term is justified in terms of encouraging investment in AI-generated works and technology.

We believe that the current period of protection should be maintained.

4. What are your views of the implications of the policy options and of AI technology for the designs system?

No comment.

5. For each option, what are your views on the risk that AI generated works may be falsely attributed to a person?

If the current definition of authorship of a CGW is maintained (‘the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken’) then there is little reason for AI works to be falsely attributed to a person. Given our comments above on the nascent state of AI-authored works, we feel that we need to see how the field develops and the issues that arise before we can make any concrete statements about the relationship between human and AI authors, and what authorship may actually mean in terms of CGWs.

Copyright – text and data mining (TDM)

6. If you license works for TDM, or purchase such licences, can you provide information on the costs and benefits of these? For example, availability, price-point, whether additional services are included or available, number and types of works covered by the licence etc.

The publishing industry has developed a functional licensing market in tandem with AI/TDM developers that delivers high-quality content for TDM and AI purposes. This should be encouraged and incentivized through non-legislative means. ALPSP members have been active in licensing their content for TDM for many years and TDM/AI/machine reading licensing is a growing and increasingly significant revenue stream. For reasons of commercial sensitivity, it is difficult to precisely quantify the value of this market overall and by territory, and most current estimates of market size are likely to be substantially under-called given the significant growth in activity in recent years.

Our members license their publications for a variety of TDM activities, across multiple market sectors, and to all types of organisations. Some examples of the types of companies and sectors where our members either have licensing arrangements for TDM specifically for AI purposes, or have had serious discussions, include:

• Consultancies developing information analytics tools using AI, or conducting specific research projects, usually funded by large pharmaceutical or chemical companies
• Major technology companies and start-ups developing AI service layers or industry-specific tools for commercial exploitation
• Start-ups building AI data analytics tools for the pharma and health sciences industry
• Companies using AI to conduct primary research in pharma/life sciences
• Academic research into AI tools for teaching and education

Each of these activities is likely to require different licensing terms and models, and our members have developed new and innovative licences that meet the wide-ranging needs of TDM customers, from high-value commercial TDM activity through to academic project work. There are huge differences in the types of licence requests our members receive, and our members continue to develop new licensing models to support market needs across the TDM/AI/machine reading sector.

Licences are usually granted to cover all published material, or relevant segments of content (e.g., specific subject areas or journals). Our members offer flexible and bespoke licences that enable customers to license the specific content they need for their purposes. A wide variety of licence types are available, but the most common are either standalone (the customer purchases data for TDM/AI/machine reading purposes only), or as an add-on to a subscription licence.

We should be clear that, as we represent a huge variety of publishers in terms of scale, subject areas, and technical capabilities, there is a wide range of experience of TDM licensing and business models, but nearly all our members are engaged in some form of licensing activity and working to develop flexible and practical models to enable their content to be used for TDM/AI/machine reading purposes.

It is a commonly held misconception that TDM is the same thing as crawling publisher platforms or the open web to obtain content. This is not the case. Our members have made significant investments in content delivery services to enable discovery and delivery of content to customers for TDM purposes. Some publishers permit crawling of their websites to access content, but this is a ‘brittle’ process, prone to error and can have significant negative impacts on the performance and stability of publisher platforms for human readers. The majority of publishers will deliver content directly to customers, for example via FTP or Amazon S3, or have developed sophisticated search and delivery APIs so customers can find and access content at scale in a flexible manner.

We cannot comment on pricing as we are a trade association.

7. Is there a specific approach the government should adopt in relation to licensing?

Licensing is a flexible approach that reduces friction between rightsholders and users and can be adapted easily to suit different sectors and situations. A licensing regime recognises the value of content provided by rightsholders to TDM activities in support of the creation of robust AI systems and provides clarity for all actors in terms of permitted actions. The current bilateral licensing approach works well in practice and enables customers to access high-quality content under clear terms and conditions.

There are some areas which it might be useful for the government to review in terms of licensing that could improve the licensing framework for some actors and market sectors:

• Collective licensing: it would be useful to explore whether a collective licensing solution might help the ‘long tail’ of smaller publishers license their materials at scale for TDM
• Model licence terms: to ensure that there is a level playing field, and clear understanding among all actors of what TDM means in practice and how it can be licensed, it may be worthwhile to develop model licence clauses that could be adopted as needed to cover TDM/AI/machine reading rights and obligations. One example of this is the model licence clause developed by the STM Association and the Pharma Documentation Ring, which was used to extend subscription licences to cover TDM under mutually agreed terms
• Pilot licences: it is clear that some sectors feel (incorrectly) that they are excluded from current licensing opportunities. Developing voluntary pilot licensing schemes specifically for these sectors could help build a more functional market on clear and agreed terms. ALPSP and its members are keen to engage with all stakeholders in developing workable licensing solutions for the reuse of copyright protected content.

We believe that good faith engagement between customers and users in the areas described above could expand and extend an already-functioning licensing market in positive ways. Such models should not however be used to reduce rightsholders' freedom to contract or decide the terms and conditions on which content and services are made available to customers and users.

8. Please rank the options in order of preference (most to least preferred) and explain why.

Our preferred option is #1 (improve licensing environment for the purposes of TDM), for the reasons already outlined. We believe that the other options, in particular #2 and #4, would cause significant harm to our members’ interests and would clearly prevent normal exploitation of the works that they publish, given that there is a healthy and growing licensing market in place for TDM that supports innovation in AI and machine learning.

As noted above, option #1 would provide the opportunity for stakeholders in the TDM market to work together in open and constructive dialogue with the IPO to develop licensing best practice. It would be helpful for the IPO to share market data on gaps and frictions in the existing market, as perceived by some market actors and segments.

Our next preference would be option #0 (make no legal change), which would effectively preserve the status quo. At present, rightsholders support TDM for academic and research purposes under the current exception, and as described previously have developed a functional and flexible licensing market that shows no need for further legislative intervention at the present time.

Our third preference is for option #3 (adopt a TDM exception for any use, with a rightsholder opt-out), on the assumption that this would be framed in a manner similar to Article 4 of the EU Copyright Directive. In theory, Article 4 provides TDM users with a framework for identifying content that is available for TDM and for clear visibility as to where licences are required for TDM. We have some concerns with how Article 4 can be implemented from a technical perspective (as described in question 9). In addition, ALPSP represents numerous smaller publishers who may find it difficult to implement opt-out provisions from a technical perspective, so any exception providing an opt-out would need to be clearly framed and permit opting-out through licence terms and platform-level declarations rather than being solely ‘machine-readable.’ See further comments below.

Options #2 and #4 would cause significant harm to our members’ interests and result in the disappearance of the TDM/AI/machine reading licensing market. This would benefit large technology firms (often based overseas) at the expense of UK-based content providers and rightsholders.

9. If you have experience of the EU exception with opt out for rights holders, how has this affected you?

Our members have explored implementing the Article 4 opt-out in two ways: first, through language in licensing agreements, and product/platform-level terms and conditions/terms of service. Second, through participation in a w3c working group which is developing a ‘machine-readable’ protocol that could be used by publishers to communicate the reservation of rights. As far as we are aware, no members have yet implemented the latter as the protocol is still under development and would need to be adopted by a broad spectrum of rightsholders and (especially) users in order to be worthwhile.

The main issue with the ‘opt-out’ is that the language in the Copyright Directive is unclear, and the idea of a machine-readable opt-out conflates two separate processes: accessing content for TDM (which can be done in a variety of ways, one of which is crawling the web and/or publisher platforms) and the activity of TDM itself (which takes place offline). If the government is seriously considering option #3, then we suggest that it needs to provide for stakeholder dialogues to refine and further develop what a ‘machine readable’ opt-out actually means in practice.

10.  How would any of the exception options positively or negatively affect you? Please quantify this if possible.

We believe there is a functioning licensing market in the UK for TDM/AI/machine reading, which does not require legislative intervention at this point in time. Licences are flexible tools which enable users and customers to access high-quality content for TDM and AI purposes. Legislative intervention in the form of further copyright exceptions would cause significant harm to the current licensing market, increase rather than decrease market friction, and enable a wholesale transfer of value to the technology sector.

Any broadened TDM exception (such as options #2 or #4) would have a significant impact on publishers’ and rightsholders’ ability to continue to innovate as the TDM/AI industry evolves, as well as on their ability to monetise their content in this fast-growing field. Such changes would prevent the development of business models and services to support content discovery, licensing and delivery and ensure that there is in practice no functional market for TDM or AI. Given the value of the TDM licensing market, as outlined above, any policy change here would have a seriously negative impact on the sector and on innovation to support TDM. Any policy intervention of this nature would send a dangerous message that the key issue for legislation is the reduction of content acquisition costs for a vocal minority of stakeholders, rather than spurring innovation in the AI sector.

Specifically, ALPSP has significant concerns around:

• the impact on platform security and stability if users think they can crawl rightsholders’ platforms without restriction in order to access content
• unauthorised redistribution of content if there is no license in place to cover basic terms and conditions of access to high-value subscription or purchased content
• whether any exception would be sufficiently technology-neutral to accommodate changes and developments in an innovative and evolving market
• the limitation of any exception’s ability to accommodate rightsholders’ and creators’ concerns regarding ethical obligations, accountability and transparency that apply to the downstream re-use of scientific and academic content
• potential harm to publishers by preventing the normal exploitation of published work (as outlined by the Berne three-step test) and encouraging the creation of products and services that could substitute for those created and offered in the market by publishers and rightsholders

ALPSP members are at the forefront of developing innovative licensing models for AI/TDM/machine reading and would welcome the opportunity to work with the IPO and other stakeholders to improve the licensing environment in the UK to the benefit of all involved in the sector.

Patents

11. Please rank these options in order of preference (most to least preferred) and explain why?

12. Would the changes proposed under Options 1, 2 and 3 have any consequential effects on the patent system, for example on other patentability criteria?

For options 1 and 2:
13.  If UK patents were to protect AI-devised inventions, how should the inventor be identified, and who should be the patent owner? What effects does this have on incentivising and rewarding AI-devised inventions?

14. In considering the differences between options 1 and 2, how important is it that the use of AI to devise inventions is transparent in the patent system?

15. Would the UK adopting option 2 affect your global patent filing strategy, if so, how?

For option 3:
16. What term and scope of protection should a new right offer?

17. What should the criteria for grant of a new right be and why? Particularly should it:
a) Replicate the current requirements for a patent?
b) Set a different bar for inventive step?
c) Be an automatic or registered right?

General

18. What role does the IP system play in the decision of firms to invest in AI?

19. Does the first mover advantage and winner-take-all effect prevail in industries adopting AI? How would this affect the impact of the policy options proposed on innovation and competition?

20. How does AI adoption by firms affect the economy? Does the use of AI in R&D lead to a higher productivity?

21. Do the proposed policy options have an impact on civil society organisations? If so, what types of impacts?

Section B: Respondent information

A:  Please give your name (name of individual, business or organisation).

Wayne Sime
Chief Executive
The Association of Learned and Professional Society Publishers (ALPSP)

B: Are you responding as an individual, business or on behalf of an organisation?

1) Business – please provide the name of your business

2) Organisation – please provide the name of the organisation

The Association of Learned and Professional Society Publishers (ALPSP)

3) Individual – please provide your name

C: If you are a responding on behalf of an organisation, please give a summary of who you represent.

ALPSP is the international trade association which supports and represents organisations that publish scholarly and professional content. It has nearly 300 member organisations across 30 countries. Its vision is to foster a culture that will strengthen and advance academic and scholarly communications in all its forms. It does this by:

• Leading and enabling knowledge-sharing across its membership and the wider community
• Fostering the creation of international connections through networking and collaboration
• Informing and influencing policy and decision-making, advocating for society, not-for-profit and community-led publishing
• Promoting inclusivity and diversity, removing barriers to participation and increasing transparency

D:  If you are an individual, are you?

1) General public

2) An academic

3) A law professional

4) A professional in another sector – please specify

5) Other – please specify

E: If you are responding on behalf of an organisation, are you?

1) An academic institution

2) An industry body

3) A licensing body

4) A rights holder organisation

5) Any other type of organisation - please specify

F: If you are responding on behalf of a business or organisation, in which sector(s) do you operate? (choose all that apply)

1) Agriculture, forestry and fishing

2) Mining and quarrying

3) Manufacturing – Pharmaceutical products

4) Manufacturing – Computer, electronic and optical products

5) Manufacturing – Electrical equipment

6) Manufacturing – Transport equipment

7) Other manufacturing

8) Construction

9) Wholesale and retail trade; repair of motor vehicles and motorcycles

10) Transportation and storage

11) Information and communication – Publishing, audio-visual and broadcasting

12) Information and communication – Telecommunication

13) Information and communication – IT and another Information Services

14) Financial and insurance activities

15) Real estate activities

16) Scientific and technical activities

17) Legal activities

18) Administrative and support service activities

19) Public administration and defence

20) Education

21) Human health and social work activities

22) Arts, entertainment and recreation

23) Other activities – please specify

G: How many people work for your business or organisation across the UK as a whole? Please estimate if you are unsure.

1) Fewer than 10 people

2) 10–49

3) 50–249

4) 250–999

5) 1,000 or more

H: The Intellectual Property Office may wish to contact you to discuss your response. Would you be happy to be contacted to discuss your response? YES

I: If you are happy to be contacted by the Intellectual Property Office, please provide a contact email address. wayne.sime@alpsp.org

J: Would you like an acknowledgement of receipt of your response? YES