AI Generated Inventions: Implications for Patent Valuation

Last week, Professor Adam Jaffe (Brandeis University, USA & Motu Economic and Policy Research, New Zealand) joined for a presentation assessing the implications on the patent system from the rise of inventions created by AI.

The subject is topical; we’ve previously covered Thaler -v- Comptroller, a case to be heard by the UK Supreme Court relating to an AI that generated two inventions and a dispute following the AI being listed as an inventor, and we’ve considered non-human inventions from a patent valuation perspective.

The talk was founded on a dual premise that: (1) AI systems soon will be (or already are) making inventions, and (2) the patent system as it stands is not ready for that. Whilst patent law in most jurisdictions does not make allowances for AI to be an inventor, and that position looks likely to hold, Professor Jaffe suggested it is both unavoidable and undesirable to seek to halt the rise of AI invention machines. A change in the law will likely be required.

It is theorised that AI would be capable of creating inventions at a very low marginal cost and in substantial number. Should those inventions be patentable, either because the law changes to accommodate AI as an inventor or because AI owners exploit the indistinct line between machine-assisted invention and machine-generated invention to list themselves as inventors, patent offices could be flooded with new applications. It is important, therefore, to discuss how AI inventions relate to the fundamental goals of the patent system. A predictable and fair patent system, and one that incentivises innovation are obviously desirable, yet it is on the goal of protecting investments that develop an invention to a marketable product or process that is particularly interesting.

In a previous article considering AI inventions, we speculated that AI inventors would lead to ‘invention factories’ – businesses who produce patentable inventions using their AI invention machine and sell or license those patents to others who exploit them to bring patents to market as products. We can therefore envisage a scenario with three elements: (1) AI inventors producing many more inventions each year than are currently created, (2) we assume patent offices are able to keep up, and so many more patents are granted on these inventions each year, and (3) those patent owners are unwilling or unable to exploit those patents and bring products to market. In this scenario, a substantial patent marketplace emerges.

This raises important questions for patent valuation. In any market, value is most simply determined by what someone else might pay for a product, and therefore an increase of supply (patents) into that marketplace depresses the value of each individual item given the increase in choice for buyers. Comparative IP valuation methods may need to adjust to account for this if the comparison is based on a different, non-AI generated, patent marketplace. Yet, each potential buyer of patents will have different capacities to exploit different types of patents – a chemical manufacturer has no use for a patent relating to the production of bread, for example. Increase in supply therefore, may not substantially depress value. We must also keep sight of the importance of a patent’s quality to patent valuation. A high-quality patent – by which we mean one with a long term of protection, substantial coverage and strong potential commercial gains when properly exploited – remains a valuable item whether created by a human or AI. Identifying those patents will be crucial for commercial success in an AI invention landscape.

Should a patent marketplace emerge as we have speculated, the demand for effective patent valuation to complement patent-related dealmaking will rise. Moreover, with so many patents potentially on offer, the importance of good patent valuation only grows. The problem for patent valuation therefore may be the same as for the patent offices – volume. The skills, however, remain the same. Patent valuation experts who can identify and understand market outlooks, growth implications and draw valuable insights from the comparative deal landscape to make well-founded, accurate and commercially useful judgements on these patents will be invaluable.

We are very grateful to Professor Jaffe for sharing his ongoing work with us. The question of AI inventions and how they fit into the patent system is an important and fluid issue. This talk demonstrated both the considerable interest in the area, and the need for open-minded thinking about how we might negotiate the problems and opportunities raised by AI inventions to achieve the greatest benefit from them.

The full webinar is available at: https://www.youtube.com/user/Oxfirst.

Further discussion article on non-human inventions from a patent valuation perspective: https://oxfirst.com/insights-&-news/thaler-v-comptroller-a-patent-valuation-perspective-on-non-human-inventions/

IPOs in the United States Drop

According to Ernst & Young and PwC, the number of IPOs and their value in the United States have dropped significantly from the similar time period from the prior year.  Interestingly, PwC notes that China's numbers of IPOs has remained relatively strong.  Cal Matters and others report that California's number of IPOs is very weak compared to the last year.  As pointed out by commentators, it seems to make sense to wait and see in this market--if you are able.  Notably, if IPOs as an exit strategy dry up, it makes acquisitions a much more important alternative.  It will be interesting to see the direction regulators in the United States take toward acquisitions--particularly in the tech space.  

Implications of the Updated IEEE IPR Policy for FRAND royalty rate determination

 

The Institute of Electrical and Electronic Engineers (IEEE) has announced a change to its IPR policy in relation to standards which will take effect from 1^st January 2023.

The changes relate to Standard Essential Patents (SEPs) which patentees agree to license to downstream innovators on a fair, reasonable and non-discriminatory (FRAND) basis. These changes hold relevance to FRAND royalty rate determinations. A press release from IEEE stated these changes “are intended to improve the clarity of IEEE’s standard processes relating to patented technologies, while offering more options for stakeholders”.

IEEE’s changes relate to the Reasonable Rate definition within the Bylaws (accessible here, p.2). They have implications for the choice of the royalty based in a FRAND royalty rate calculation and relate to the bigger pictures as to what is FRAND. Regardless of the extent to which IEEE ever recognised the smallest saleable practicing unit (SSPPU), these changes impart a carte blanche for FRAND royalty base determination to be derived from market value at the component level.

The Adequate Royalty Base in the FRAND Royalty rate valuation

The choice of a royalty base is decisive in the calculation of an appropriate FRAND royalty rate. The debate has focused on deciding whether the smallest saleable practicing unit (SSPPU) or the entire market value (EMV) is the most appropriate methodology.

And as in many cases, damages are calculated by factoring a royalty base with a royalty rate; the value of the royalty base is of substantial importance. In the SSPPU approach, the price of the component which houses the SEP technology is used as the calculation base.

On the other hand, if the conditions for EMV is met and the technology is concluded to be the primary driver of demand to the entire end-product and the primary contributor to the functionality, the entire market value approach is used.

The SSPPU methodology originates from a suit between Cornell v. Hewlett Packard Co. where the Plaintiff initially sought damages based on the end-product, but later had to change its claim as the court held the end-product as an inappropriate royalty base, concluding that calculation must be apportioned down to “the smallest salable infringing unit with close relation to the claimed invention.”

Shortly after, the same reasoning would be echoed in Lucent Technologies, Inc. v. Gateway, Inc where the court concluded that there was no evidence that the infringed feature in question drove the entire demand for the end-product and therefore a calculation with the end-product as the royalty base would be misplaced. In Lucent v. Gateway, as in Cornell v. Hewlett Packard, decisions did however leave the door open for the future usage of the end-product as a royalty rate while calculating damages. More specifically, in Lucent, the court concluded that there was ‘nothing inherently wrong’ with using the end-product given that the proportionate value of the infringed feature or component to the product is taken into account.

What the Changes are about and how they Affect the valuation of Patents

We think it is valuable to show these amendments in full, with the changes italicised. Where eliminations have been made, they are shown with a strikethrough:

““Reasonable Rate” shall mean appropriate compensation to the patent holder for the practice of an Essential Patent Claim excluding the value, if any, resulting from the inclusion of that Essential Patent Claim’s technology in the IEEE Standard. In addition, determination of such Reasonable Rates should include, but need not be limited to, the consideration of:

Some optional considerations for determination of Reasonable Rates are:

·       The value that the functionality of the claimed invention or inventive feature within the Essential Patent Claim contributes to the value of the relevant functionality of the smallest saleable Compliant Implementation that practices the Essential Patent Claim or to another appropriate value level of the Compliant Implementation.

·       The value that the Essential Patent Claim contributes to the smallest saleable Compliant Implementation or to another appropriate value level of the Compliant Implementation that practices that Essential Patent Claim, in light of the value contributed by all Essential Patent
Claims for the same IEEE Standard practiced in that Compliant Implementation.

·       Existing licenses covering use of the Essential Patent Claim, where such licenses were not obtained under the explicit or implicit threat of a Prohibitive Order, and where the circumstances and resulting licenses are otherwise sufficiently comparable to the circumstances of the contemplated license.”

The bylaws have also been amended in relation to licence negotiations (p.4). Again, the accepted amendments are shown by italics with eliminations shown with a strikethrough:

“The Submitter and the Applicant should engage in good faith negotiations (if sought by either party) without unreasonable delay or may litigate or, with the parties’ mutual agreement, arbitrate: over patent validity, enforceability, essentiality, or infringement; Reasonable Rates or other reasonable licensing terms and conditions; compensation for unpaid past royalties or a future royalty rate; any defenses or counterclaims; or any other related issues. The Submitter of an Accepted LOA who has committed to make available a license for one or more Essential Patent Claims agrees that it shall neither seek nor seek to
enforce a Prohibitive Order based on such Essential Patent Claim(s) in a jurisdiction unless the against an implementer who is willing to negotiate in good faith for a license. Seeking further information upon initial notice of infringement or choosing to litigate or arbitrate over any of the foregoing issues, however, does not by itself mean that a party so choosing is unwilling to negotiate in good faith. fails to participate in, or to comply with the outcome of, an adjudication, including an affirming first-level appellate review, if sought by any party within applicable deadlines, in that jurisdiction by one or more courts that have the authority to: determine Reasonable Rates and other reasonable terms and conditions; adjudicate patent validity, enforceability, essentiality, and infringement; award monetary damages; and resolve any defenses and counterclaims. In jurisdictions where the failure to request a Prohibitive Order in a pleading waives the right to seek a Prohibitive Order at a later time, a Submitter may conditionally plead the right to seek a Prohibitive Order to preserve its right to do so later, if and when this policy’s conditions for seeking, or seeking to enforce, a Prohibitive Order are met.”

There are three documents related to these changes: The IEEE Standards Association Board of Governors (IEEE SA BOG) approved proposed updates to the IEEE SA Standards Board Bylaws, approved proposed updates to the IEEE SA Letter of Assurance (LOA) form, and approved proposed updates to the patent policy FAQs. The changes detailed above appear in the Bylaws, but the changes are also reflected in the other two documents.

Take Away

These changes further illustrate IEEE’s desire to review its position on the bigger questions within the SEPs landscape, particularly ongoing debates relating to holdup vs holdout. For valuation professionals these updates are of importance as any FRAND royalty rate calculation needs to consider the IPR policy of the Standard Setting Organization as they govern the standards with which the patents are associated with. In particular the choice of the royalty base is an important decision to make in any FRAND royalty rate calculation. By updating its IPR policy the IEEE appears to offer further choices for the valuation of standard essential patents. How this will affect the practice of FRAND royalty rate calculation and IP valuation remains to be seen.