Nearly four years have passed since the Enlarged Board of Appeal issued the decision in G 1/19 (Pedestrian simulation). The decision addressed the patentability of computer simulations and, more broadly, software inventions. Given that artificial intelligence (AI) is classified by the European Patent Office (EPO) as a combination of software and a mathematical method, G 1/19 also has a direct impact on AI-related patent applications.
G 1/19 confirmed that the two-hurdle COMVIK approach, first established in the early 2000s, remains the correct framework for assessing the patentability of software inventions in Europe. The first hurdle, under Article 52 EPC, is relatively easy to clear – an invention must contain at least one technical feature (for example, a processor) to avoid classification as a “non-invention”. Overcoming the second hurdle under Article 56 EPC (inventive step), however, is much more involved. It must be shown that the claimed software produces a technical effect that serves a technical purpose which extends beyond the mere implementation of the software on a conventional computer system. Any claim feature that does not contribute to such a technical effect is disregarded when assessing inventive step.
G 1/19 included an extensive discussion on what constitutes a relevant technical effect. While it provided intriguing insights, its long-term influence on EPO practice was initially uncertain. Now, with over a hundred Board of Appeal decisions citing G 1/19, its impact is becoming clearer, allowing us to see how its reasoning is shaping the assessment of inventive step for AI and software inventions. A notable trend emerging in these decisions is that inventive step objections now come in a variety of ‘flavours’, each with different implications and potential resolutions. This shift seems to result from the detailed clarification of technical effect requirements in G 1/19. While classical obviousness remains the ‘vanilla’ standard, other reasoning has gained prominence, including the alleged effect not being technical, not being credible, or not being achieved across the whole claim scope. In this article, we explore these distinct flavours of inventive step objection as they arise for AI and software inventions.
Alleged effect is not technical
As mentioned, the second hurdle of the COMVIK approach requires that the claimed invention produce a ‘technical’ effect. Based on the case law citing G 1/19, the definition of ‘technical’ remains a contentious issue. This is unsurprising as the Enlarged Board in G 1/19 refrained from explicitly defining the term, leaving room for continued debate.
Longstanding exclusions from patentability, such as business and financial methods, administrative processes, and the mere presentation of information, remain unchanged. That said, there appears to have been an increase in Board of Appeal decisions recognising technical effects arising from the technical implementation of business methods. However, the volume of appeals on this issue suggests that Examining Divisions of the EPO are yet to fully align with this evolving practice.
Interestingly, applicants now seem to be facing greater challenges in demonstrating technical effects for core software and AI inventions. For instance, T 1998/22 (Wide and deep machine learning models/GOOGLE) illustrates how technical contributions for core AI are subject to heightened scrutiny. Software and AI that has an application such as healthcare or energy continues to be considered technical.
Alleged effect is not credible
A relatively recent trend is the rise of inventive step objections based on the credibility of a claimed technical effect. While G 1/19 did not explicitly refer to the need for a ‘credible’ technical effect, the Boards of Appeal have increasingly raised objections with this reasoning. Objections of this nature appear when the Board is not convinced that an alleged technical effect is actually (credibly) achieved by the claimed invention. The way to remedy this is typically through evidence (e.g. data) showing the technical effect is indeed achieved.
This reasoning bears similarities to former ‘plausibility’ case law in life science and chemical fields, which was recently overturned in G 2/21 (Reliance on a purported technical effect for inventive step (plausibility)). According to G 2/21, only technical effects encompassed by the technical teaching of an application can be relied upon for inventive step, and post-filing data may be used to substantiate such effects. T 0687/22 (Active occlusion cancellation/GN HEARING) is one of the few cases to explicitly link G 2/21 with the assessment of credible technical effects for software inventions. If this trend continues, future decisions may clarify which technical effects can be relied on, and when and how applicants should provide supporting data of technical effects for software and AI inventions.
Alleged effect is not technical and/or not credible across the whole scope of the claim
It is an established principle at the EPO that alleged effects must be achieved across substantially the whole scope of the claim. This prevents overly broad claims from being granted based on incorrect technical effects. For software inventions, this would seem to manifest as an inventive step objection where the effect is deemed not credible across the entire claim scope. Numerous Board of Appeal decisions have applied this reasoning. We only found one instance where an applicant has faced such an objection and successfully argued against it based on the content of the application as filed, T 0814/20 (Adapted Visual Vocabularies/CONDUENT).
G 1/19 reinforced this line of reasoning by explicitly stating, for the first time, that an invention must be technical across the entire claim scope. As we have seen from the case law that followed, this reasoning is often raised when a claim covers both technical and non-technical implementations, especially if the application explicitly acknowledges applicability in non-technical fields (e.g., finance). In such cases, the Boards are seen to conclude that the alleged technical effect does not apply across the entire claim scope, resulting in a lack of inventive step.
Claimed solution is obvious
The Boards’ fundamental approach to obviousness remains unchanged. The key question is whether, starting from the prior art and faced with the problem of providing the alleged technical effect, would it have been obvious to a skilled person to arrive at the claimed invention.
One notable trend, however, is the increased use of classical obviousness reasoning in cases where technical effects are acknowledged on the basis of technical implementation. The Boards frequently deem such implementations to be straightforward, arguing that non-technical (e.g. business) claim features, even if these are not found in prior art, can still form part of the alleged effect. This is different to technical claim features, which should not appear in the technical effect as this is considered to amount to hindsight.
Conclusion
Following G 1/19, several aspects of the Boards’ approach to inventive step for AI and software inventions remain unchanged. The assessment of obviousness continues to follow established principles, and the Boards’ stance on what constitutes ‘technical’ subject matter has remained relatively stable. Some developments have even been positive for applicants; the Boards appear to be more willing to recognise technical effects arising from technical implementation. That said, many such cases then go on to fail on obviousness grounds, as the technical implementation is considered routine or standard. Other aspects of establishing inventive step have become more challenging. The Boards increasingly expect applicants to establish that a claimed technical effect is credible and applicable across the entire claim scope. As a result, claims must more carefully align scope with technical effect, and supporting evidence (e.g. data) must be sought.
Our takeaway is that navigating the inventive step requirements for AI and software in Europe is becoming more complex, but also more predictable. While applicants face stricter scrutiny, recent case law arising from G 1/19 provides clearer direction on what constitutes a technical effect that will lead to an inventive step. Success will depend on aligning claims with technical contributions and ensuring supporting evidence to prove credibility of any alleged effects is available.
View our related articles here:
