Opinion from PCC on consent to use generative AI agents
Note to readers: This opinion relates to an aspect of the use of generative AI agents in the work of patent attorneys. This is a fast-changing area in which developments may render the opinion less relevant than at the time it was drafted. Members are encouraged to interpret the opinion in line with relevant developments; and/or to seek an updated opinion from the Professional Conduct Committee.
Summary of enquiry
The Enquiring Member wishes to know whether the recommendation, in Guideline 4 of the recently published Guidelines on Use of Generative AI in the Work of Patent Attorneys (herein “Guidelines”), requires the obtaining of specific consent from a client to the use of generative AI when drafting text summarising existing prior art in a patent application or response.
Relevant Provisions
Guideline 4 of the aforementioned Guidelines; and the Guidelines generally
“Guideline 4: Members must in all instances establish, in advance of using generative AI in their cases, the wishes of their clients with regard to the use of generative AI.”
Opinion
General Principles Underlying Guideline 4
As a preliminary remark, it is important to keep in mind that the Guidelines are neither Rules of Professional Conduct nor formal recommendations on conduct of epi members. Rather, they are intended to assist members to judge what are appropriate practices in an area of activity that is recognised as being contentious and imprecise.
Hence there is no directly applicable sanction for any failure to observe the provisions of the Guidelines. However the Disciplinary Bodies of epi and the EPO may interpret non-observance of the Guidelines as evidence that a European Patent Attorney may have breached for example an Article of the Regulation on Discipline.
Guideline 4 is intended to assist Members of epi to avoid liabilities that may result from any failure to understand the wishes of a client concerning the use of generative AI tools. Guideline 4 urges Members to establish the wishes of their clients with regard to the use of generative AI tools; but it does not prescribe how, or how frequently, such wishes are to be identified or recorded.
In view of this it firstly is clear that any form of establishing the wishes of a client is likely to comply with the Guideline, as long as it achieves the objective of allowing the Member clearly to know whether in given circumstances the use of AI tools is accepted by the client.
Obviously a written form of consent is easier to recall and refer to than for example verbal consent; but whether a written record of consent to use generative AI (or indeed a prohibition on its use) is created is a matter of practical organisation for the European Patent Attorney and the client.
Secondly the Guideline does not call for the obtaining of consent more than once, although as explained below in some situations this may be desirable.
In particular in this regard if the areas of endeavour of the client alter over time depending on the nature of the change it may be prudent not to assume that consent given in the past continues to be valid. In such a situation it would be desirable to assess whether previously given consent remains applicable; and if necessary to obtain renewed consent.
Additionally, developments in the capabilities of AI tools and, somewhat distinctly, changes in the ownership of the client entity may make it desirable to update any consent previously obtained.
Such aspects are matters of risk evaluation, and if in the professional judgement of the Member an existing consent remains valid (or on the balance of probabilities is likely to remain valid) there is no need repeatedly to update the consent. Indeed, unnecessarily asking for consent updates may irritate clients and thereby harm the reputation of the European Patent Attorneys’ profession.
It follows that a European Patent Attorney in obtaining consent to use generative AI tools should take the minimum steps that are consistent with achieving a clear understanding within the meaning of Guideline 4.
As a practical matter a private practice attorney probably can satisfactorily comply with Guideline 4 by including consent provisions in for example correspondence (such as terms of engagement, or an engagement letter) exchanged at the commencement of a working relationship with a client. In this case, the principal areas and the extent in which generative AI is to be used should be indicated.
It may also be possible to comply with Guideline 4 through the posting of a statement on a firm’s website. In particular a statement to the effect that a firm intends to use generative AI tools unless the client indicates a contrary wish in the view of the author would comply with Guideline 4.
In either of these cases any indication of an intention to use generative AI tools should not be anything other than clear and readily available to the client. Obscure references to consent, or references that are not prominent in a firm’s website, are unlikely to comply with Guideline 4.
European Patent Attorneys working in-house should interpret the foregoing mutatis mutandis with regard to their precise circumstances. It seems very likely that many internal policy statements will give rise to compliance with Guideline 4.
A further overarching principle is that a Member must be alert to possible misunderstandings of the operation of AI tools by clients. In particular, Members should be alert to signs of clients not being aware of the risks of loss of confidentiality of information through use of generative AI tools.
Specific Aspects of the Enquiry
The enquiry is right to imply that statements of prior art in patent specifications are in a different category than, for example, patent claims which must be novel and hence cannot usually be prepared using anything other than closed AI systems (which for the most part are not available to patent attorneys).
It is reasonable to conclude that, assuming the client consent explained above has been obtained, statements of prior art in specifications may be prepared using AI tools.
Consent to use AI tools for this purpose is likely to be long-lasting or perhaps even (in practice) perpetual. This is because the drafting of statements of prior art is an activity unlikely to give rise to circumstances creating a need to renew any usage consent given, unless there is a change of ownership of the client entity.
Statements of prior art in responses require more care, since the choice of prior art references relied on and the emphases given to them are likely to reflect the instructions of the client. These of course are confidential and, in most situations, legally privileged. The patent attorney is not at liberty to place this confidentiality or privilege at risk, even if consent to use AI tools has clearly been provided. When using AI tools in the preparation of responses therefore Members must remain constantly alert to the risk that one may inadvertently release confidential information that may harm the client.
It should of course be repeated that compliance with the Guidelines is not mandatory, although it is recommended. The Guidelines should be viewed independently of any regulatory obligations, and no aspect of complying with the Guidelines may substitute for compliance with any other obligations attaching to the work of a European Patent Attorney.
Summary of Opinion
Guideline 4 is intended to assist patent attorneys to avoid liabilities that could arise through the use of generative AI tools without the consent of clients. Compliance with Guideline 4 is not mandatory, although failure to comply may have implications in for example cases that may be heard by the disciplinary bodies of epi and the EPO.
The requirement in Guideline 4 to learn the wishes of clients concerning the use of AI tools is not intended to create an onerous work burden; and indeed excessive seeking of consent is not recommended.
With the recommended client consent, drafting of prior art statements seems to be permissible, under Guideline 4, in patent specifications.
Drafting of prior art statements also is permissible in responses but Members must be alert to the chance that AI prompts selecting or emphasising particular prior art documents (or combinations of documents) may breach the confidentiality and/or privilege of clients. Hence care must be exercised when using AI tools to indicate prior art in responses.
This opinion does not have regulatory force and is prepared with the intention to provide helpful assistance. An opinion provided in accordance with Article 7(c) CoC is not binding on the Disciplinary Bodies. No liability of any kind attaches to epi, its Professional Conduct Committee or any members of that Committee in respect of this opinion.
