The European Commission report Trends and Developments in Artificial Intelligence: Challenges to the IPR Framework by Christian Hartmann, Jacqueline Allan, P. Bernt Hugenholtz, Joao Pedro Quintais and Daniel Gervais states
This Report examines the state of the art of copyright and patent protection in Europe for AI- assisted outputs in general and in three priority domains: science (in particular meteorology), media (journalism), and pharmaceutical research. “AI-assisted outputs” are meant as including productions or applications generated by or with the assistance of AI systems, tools or techniques.
As the state of the art reviewed demonstrates, the use of AI systems in the realms of culture, innovation and science has grown spectacularly in recent years and will continue to do so. AI systems have become almost ubiquitous in meteorology and in pharmaceutical research and are making deep inroads into media and journalism. Outside these distinct domains, AI systems are being used to generate diverse literary and artistic content, including translations, poems, scripts, novels, photos, paintings, etc. Likewise, a wide variety of innovative and inventive activity relies on AI systems for its development and deployment, from facial recognition to autonomous driving.
While AI systems have become – and will become – increasingly sophisticated and autonomous, this Report nonetheless assumes that fully autonomous creation or invention by AI does not exist, and will not exist for the foreseeable future. We therefore view AI systems primarily as tools in the hands of human operators. For this reason, we do not enquire whether AI systems should one day be accorded authorship or inventorship status under future IP Law. We also do not examine the IP protection of AI systems per se; legal issues concerning the input of protected subject matter into AI systems (e.g. for text-and-data mining); nor algorithmic moderation or enforcement of IP subject matter, as these topics are beyond the scope of this analysis. Analysis of legal protection regimes beyond copyright and patent law (e.g. trade secrets, unfair competition and contract law) is also outside the terms of reference. An important trend that does emerge from the state of the art review is that more and more AI capability is being offered “as a service” rather than as “bespoke” (tailored) AI systems. Consequently, the emphasis of our analysis is on the users (operators) of AI systems, rather than on their developers.
This Report provides an assessment of the state of the art of uses of AI in the three priority domains and a legal analysis of how IP laws currently apply to AI-assisted creative and innovative outputs. The Report concludes with recommendations regarding possible revision of IP law at the European level.
State of the Art in Uses of AI
There is no universally accepted definition of AI. At a high level, it can be defined as “computer- based systems that are developed to mimic human behaviour” or a “discipline of computer science that is aimed at developing machines and systems that can carry out tasks considered to require human intelligence, with limited or no human intervention.”
In pharmaceutical research, AI is finding patterns within large data sets and helping to automate the search process. Based mostly on machine learning, AI is assisting in disease diagnoses, predictions of drug efficacy and identification of drug characteristics (e.g. toxicity). Neural networks enable compound discovery, personalised medicine and drug repurposing. AI is being applied in finding molecular drug targets (e.g. proteins, nucleic acids) by searching through libraries of candidates, accelerating the high throughput screening needed to find a candidate substance for further investigation in drug development. It is helping in repurposing of drugs to meet new or different need, in polypharmacology (where a disease is due to multiple malfunctions of the body) and to find and accelerate the development of vaccines (by both gene sequencing and simulations of vaccines). In these processes, some measure of human intervention is usually required, either at the start or throughout the entire process, with human feedback optimising the steps.
In recent years, there has been increasing cooperation between the pharmaceutical industry and AI companies. Some companies pursue an active IP strategy and file patents in both the domains of pharmaceutical and AI technology while others sell services confidentially to pharmaceutic companies.
In the area of science, the Report examines meteorology as one of the main application areas where AI is already routinely used. Meteorology predicts the state of the atmosphere, at a certain time in a certain place or over a specific area, based on historical data and knowledge of climate and the atmosphere. Automated tasks include post-processing of weather data; predictive analytics for future forecasts; bias correction of meteorological observations; parameterisation of models to correct for radiation, turbulence, cloud microphysics, etc.; data assimilation; and local downscaling of model outputs to improve predictions.
Weather forecasts rely on vast quantities of data. The ability of machine learning to extract knowledge from complex and extensive databases makes it particularly suitable for numerical weather forecasting. Some companies support media companies in weather reporting and forecasting, providing high-precision, precise weather forecasts in multiple formats daily, including recordings, to suit the various reporting media.
In journalism, AI enables automated aggregation, production and distribution of content (data, text, images, audio or video). Assistive technologies support journalists in the creation of media content, including speech recognition, information extraction, clustering, summarising, and machine translation capabilities for multi-lingual access to sources. Generative technologies produce media content with human intervention limited to inputting the data set, defining output specifications, and quality control. Distributing technologies encompass the publication or other forms of communication (e.g. through chat bots) of automatically created content with the help of algorithms.
Several companies currently offer technologies for automated content creation for uses in diverse areas including describing product, summarising patient notes in hospitals, reporting on sports events, reporting share prices and customising local information on property markets. Other companies have in-house capabilities for automated news generation. Know-how is commonly protected through licensing models, rather than asserting ownership of IP. It also relies on the tacit knowledge within a company, with the knowledge on how to develop customer-specific systems acting as a high barrier to competitors looking to enter the market.
Legal Analysis under EU Copyright and Patent Laws
The legal analysis examines whether, and to what extent, AI-assisted outputs are protected by European copyright law, related rights or patent law. For copyright, the analysis is concentrated on the so-called EU copyright acquis and its interpretation by the Court of Justice of the EU (CJEU). The patent analysis concentrates on the European Patent Convention (EPC).
EU Copyright Law
As our inquiry into EU copyright law reveals, four interrelated criteria are to be met for an AI- assisted output to qualify as a protected “work”: the output is (1) a “production in the literary, scientific or artistic domain”; (2) the product of human intellectual effort; and (3) the result of creative choices that are (4) “expressed” in the output. Whether the first step is established EU law is however uncertain. Since most AI artefacts belong to the “literary, scientific or artistic domain” anyway, and are the result of at least some “human intellectual effort” (however remote), in practice the focus of the copyright analysis is on steps 3 and 4.
Based on a thorough analysis of the CJEU’s case law, and in light of the findings of two experts workshops, we conclude that the core issue is whether the AI-assisted output is the result of human creative choices that are “expressed” in the output. In line with the CJEU’s reasoning in the Painer case, we distinguish three distinct phases of the creative process in AI-assisted production: “conception” (design and specifications), “execution” (producing draft versions) and “redaction” (editing, finalisation). While AI systems play a dominant role at the execution phase, the role of human authors at the conception stage often remains essential. Moreover, in many instances, human beings will also be in charge of the redaction stage. Depending on the facts of the case, this will allow human beings sufficient creative choice. Assuming these choices are expressed in the final AI-assisted output, the output will then qualify as a copyright- protected work. By contrast, if an AI system is programmed to automatically execute content without the output being conceived or redacted by a person exercising creative choices, there will be no work.
Due to the “black box” nature of some AI systems, persons in charge of the conception phase will sometimes not be able to precisely predict or explain the outcome of the execution phase. This however need not present an obstacle to the “work” status of the final output, assuming that such output stays within the ambit of the person’s general authorial intent.
Authorship status will be accorded to the person or persons that have creatively contributed to the output. In most cases, this will be the user of the AI system, not the AI system developer, unless collaboration between the developer and user on a specific AI production indicates co- authorship. If “off-the-shelf” AI systems are used to create content, co-authorship claims by AI developers will also be unlikely for commercial reasons, since AI developers will normally not want to burden customers with downstream copyright claims.
A problem that might arise is the possibility of falsely claiming authorship in respect of AI productions that do not qualify as “works” for lack of human creativity. Producers or publishers might be tempted to falsely attribute authorship in such productions in order to benefit from the authorship presumptions granted under EU law, which allow the person whose name is mentioned as an author to initiate infringement procedures.
British and Irish copyright law accord authorship statu s to persons undertaking the arrangements necessary for creating computer-generated works in cases where no (human) author can be identified. These provisions have been criticised as being incompatible with EU copyright standards, since “authorless” works do not meet the EU standard of “the author’s own intellectual creation”. Perhaps, they are therefore better understood as a species of related rights.
The related rights harmonised under the EU acquis offer various possibilities for protecting AI- assisted outputs that do not qualify for copyright protection. In light of the general absence in related rights’ law of a requirement of human authorship or originality, and its rationale of rewarding economic or entrepreneurial activity, related rights will accommodate AI-assisted output in cases of insufficient human creative input.
While AI-assisted outputs in the form of aural signals (audio data) may benefit from the phonographic right, audio-visual outputs will qualify for protection under the film producer’s right. Moreover, AI-assisted broadcasts may find protection under the related rights of broadcasters. None of these related rights provide for a threshold requirement, making these regimes available for AI-assisted outputs that are generated without any creative human involvement – even absent significant economic investment. In most cases the user, not the developer, of the AI system will be deemed the owner of the related right, since it is the user that triggers the acts that give rise to these rights through his use of the AI system and output generation.
Additionally, AI-generated databases will qualify for sui generis protection under the EU Database Directive if the databases are the result of substantive investment. This includes investment in AI technology and know-how applied in producing the database. In light of the broad legal notion of “database”, the sui generis right potentially offers protection to a wide range of AI-assisted productions. However, it is currently uncertain whether investment in machine-generating data – for example, the generation of weather data with the aid of AI – may be factored in. In any case, the prerequisite of a “database” rules out protection of raw data.
Illustrated via case studies in the Report, it is impossible to make general assessments of the copyright status of AI-assisted outputs in individual cases. In some cases, where the creative role of human beings is evident at various stages of the creative process, such as The Next Rembrandt project, the output will most likely be copyright protected. In other cases, where it is difficult or even impossible to identify creative choices, such as automatically-generated sports reports or AI-assisted weather forecasts, copyright protection will be less likely. Note however that this is the same for sports reports and weather forecasts produced without machine assistance. Nevertheless, producers of “authorless” AI-assisted outputs might still find recourse in related (neighbouring) rights.
“Authorless” AI-assisted outputs will remain completely unprotected only in cases where no related right or sui generis right is available. Since such rights attach primarily to aural and audio-visual signals, as well as to databases, such cases are most likely to occur if the AI- assisted output is in alphanumerical form. Whether this absence of IP protection might justify regulatory intervention, is primarily an economic question that cannot be addressed in the context of this Report. Such intervention is justified only if no alternative protection (e.g., under trade secret protection, unfair competition or contract law) is available, and solid empirical economic analysis reveals that the absence of protection harms overall economic welfare in the EU.
Our analysis for EU copyright law and AI leads to the following conclusions and recommendations:
• Current EU copyright rules are generally sufficiently flexible to deal with the challenges posed by AI-assisted outputs.
• The absence of (fully) harmonised rules of authorship and copyright ownership has led to divergent solutions in national law of distinct Member States in respect of AI-assisted works, which might justify a harmonisation initiative.
• Further research into the risks of false authorship attributions by publishers of “work-like” but “authorless” AI productions, seen in the light of the general authorship presumption in art. 5 of the Enforcement Directive, should be considered.
• Related rights regimes in the EU potentially extend to “authorless” AI productions in a variety of sectors: audio recording, broadcasting, audivisual recording, and news. In addition, the sui generis database right may offer protection to AI-produced databases that are the result of substantial investment.
• The creation/obtaining distinction in the sui generis right is a cause of legal uncertainty regarding the status of machine-generated data that could justify revision or clarification of the EU Database Directive.
EU Patent Law
Our analysis of European patent law – and in particular the EPC – demonstrates that the requirement that an inventor be named on a patent application means that one or several human inventors must be identified. Under the EPC regime, this is essentially a formal requirement. The EPO does not resolve disputes regarding substantive entitlement, which is an issue that is governed by national law. Following this approach, the EPO decided two cases in 2020 (currently under appeal) where it considered that, because AI systems do not have legal personality, they cannot be named inventors on a patent application.
A human inventor typically has the right to be named on the application. Beyond this, inventorship and co-ownership are mostly a matter for national law. It should be noted, however, that as AI technology stands today, the possibility that an AI system would invent in a way that is not causally related to one or more human inventors (e.g. the programmer, the trainer, the user, or a combination thereof) seems remote. As technology stands, no immediate action appears to be required on the issue of inventorship at EPC level.
As regards ownership, there are at least three possible (sets of) claimants to an AI-assisted invention: the programmer or developer of the AI system; the owner of the system; and the authorised user of the system (who provided it with training data or otherwise supervised its training). Neither international law nor the EPC provide clear rules on how ownership of patents may be affected by this new type of AI-assisted inventive activity. It is therefore a matter for national laws. However, that might not require harmonisation as there does not seem to be a problem in establishing a sufficient connection between an AI-assisted invention and a patent applicant.
The granting of a patent requires that, as of the date of filing, the invention must be new (novel) and involve an inventive step. While the increasing use of AI systems for inventive purposes does not require material changes to these core concepts, it may have practical consequences for patent offices. AI systems enable qualitatively or quantitively different novelty (prior art) searches, and the practical application of inventiveness may change as certain claimed inventions may be “obvious” to a person of ordinary skill in the art (“POSITA”) due to the increasing use of AI systems. Any future changes will likely emerge in legal decisions at European (EPO Boards of Appeal) or national levels where patents will either be upheld or not.
A patent application must also sufficiently disclose the invention. The “black box” nature of some AI systems may present challenges to this requirement. In that regard, it has been suggested that a mechanism to deposit AI algorithms be established, akin to that for microorganisms (the Budapest Treaty). Although it is as yet unclear that a deposit system for AI algorithms would be useful, it seems advisable to at least consider the possibility of requiring applicants to provide this type of information, while maintaining sufficient safeguards to protect all confidential information to the extent it is required under EU or international rules.
Finally, inventions that might otherwise be patentable might be protectable as trade secrets under the 2016 Trade Secrets Directive, a topic for future study as it is outside the scope of the current work.
Our analysis for EU patent law and AI leads to the following conclusions and recommendations:
• The EPC is currently suitable to address the challenges posed by AI technologies in the context of AI-assisted inventions or outputs.
• When assessing novelty, IPOs and the EPO should consider investing in maintaining a level of technical capability that matches the technology available to sophisticated patent applicants.
• When assessing the inventive step, it may be advisable to update the EPO examination guidelines to adjust the definition of the POSITA and secondary indicia as to track developments in AI-assisted inventions or outputs.
• When assessing sufficiency of disclosure, it would be useful to study the feasibility and usefulness of a deposit system for AI algorithms and/or training data and models that would require applicants in appropriate cases to provide information that is relevant to meet this legal requirement.
• For the remaining potential challenges identified in this report arising out of AI-assisted inventions or outputs it may be good policy to wait for cases to emerge to identify actual issues that require a regulatory response, if any.
• Further study on the role of alternative IP regimes to protect AI-assisted outputs, such as trade secret protection, unfair competition and contract law, should be encouraged.
