What the 2026 AI and IP Forum concluded on AI patents

The message from the day was clear. AI is no longer a future consideration for IP teams; it is an operational reality. And the profession is grappling, in real time, with what that means for patent law, copyright, data sovereignty, and the daily work of practitioners.

Key Takeaways

• AI inventorship is settled: only natural persons qualify in the US, UK, Germany, and at the EPO.
• Inventive step is becoming the new battleground for AI patents.
• All three major offices (USPTO, EPO, UKIPO) are deploying AI inside examination workflows.
• 85% of IP professionals now use AI in some capacity (Questel 2025).
• Data sovereignty and human-in-the-loop control are now baseline requirements for AI deployed in IP workflows, not optional features.

How AI capabilities are outpacing IP law and governance

The forum opened with a stark framing from moderator C. Gregory Gramenopoulos, drawing on the 2026 Stanford HAI AI Index Report. AI capabilities continue to evolve at extraordinary velocity. Organisations are no longer experimenting; they are integrating AI into core operations. And yet, as the opening presentation made clear, the frameworks for governance, regulation, and education are struggling to keep pace.

This is something we see first-hand at Solve Intelligence. The practitioners we work with are not short of legal expertise. They are short of time. Prior art volume is rising. Prosecution is getting more complex across jurisdictions. R&D is moving faster. The bottleneck is no longer knowledge; it's capacity. That is the gap purpose-built AI is designed to close.

Patenting AI: inventorship, eligibility, and the new battleground

A heavyweight panel tackled the state of patenting AI inventions across the US and Europe. Speakers included the EPO's Chief Technology Officer Angel Aledo Lopez, the chair of EPO Board of Appeal Dr. Kemal Bengi-Akyürek, UKIPO Divisional Director Julyan Elbro, and practitioners Alexander Korenberg and Dr. Claudia Schwarz.

AI inventorship: the settled position across jurisdictions

On inventorship, the position is now settled across jurisdictions: AI cannot be named as an inventor. The DABUS saga has concluded in the US, UK, Germany, and at the EPO. The outcome is consistent: only natural persons qualify. The USPTO's November 2025 memo reaffirmed that AI is a tool, and under US law, conception remains “the formation in the inventor's mind of a definite, permanent, and operative idea.” 

The practical question has shifted to how companies manage inventorship documentation when AI plays an increasing role in R&D.

Patent eligibility for AI inventions in the US, EPO, and UK

On eligibility, the US landscape is evolving. The USPTO's August 2025 memo and the precedential Ex Parte Desjardins decision are pushing practitioners to focus on articulating the “technological improvements” of AI inventions, both in the specification and in the claims. At the EPO, the April 2026 Guidelines update addresses AI/ML inventions directly, and the panel explored what is needed to demonstrate technical character. 

In the UK, the Supreme Court's Emotional Perception decision (February 2026) has reshaped the UKIPO's approach to computer-implemented inventions, with interesting questions about convergence with EPO practice.

Inventive step: why the specification now needs to do more work

Perhaps the most forward-looking discussion concerned inventive step. Several panellists suggested this is becoming the new battleground for AI patents, particularly in light of the US Federal Circuit's Recentive Analytics (2025) decision. As AI tools become more capable, demonstrating that a specific technical implementation is non-obvious will require increasingly rigorous evidence. 

For practitioners, this raises the bar on how inventions are described from the outset. The specification needs to do more work, earlier, to establish why a particular approach is not the predictable output of a capable AI system.

Patent offices are deploying AI too

It's not just applicants using AI. All three major offices represented at the conference (the USPTO, EPO, and UKIPO) are rapidly integrating AI into examination workflows. Tools span AI-assisted prior art search such as the USPTO's ASAP! pilot and the EPO's ANSERA system, alongside AI-supported allocation, translation, and office action drafting.

The panel also raised two emerging issues that will shape prosecution strategy: how the concept of a “person of ordinary skill in the art” (POSITA) evolves as AI becomes a standard research tool, and how offices will handle the anticipated surge in AI-generated prior art.

Both questions have immediate practical implications. If the notional skilled person is assumed to have AI tools at their disposal, the threshold for inventive step rises. And if AI-generated disclosures flood the prior art landscape, efficient search and relevance assessment become essential. These are the pressures that led us to build Solve'Intelligence’s prosecution and search tools around structured, patent-specific workflows.

AI in corporate IP departments: from experimentation to operations

The panel on AI in Corporate IP Departments, moderated by Stephan Ising and featuring IP leaders from Merck, BASF, Sulzer, and thyssenkrupp, made clear that adoption has moved from experimentation to daily operational use. 

The Questel 2025 IP Outlook Report found that 85% of IP professionals now use AI in some capacity, with use cases spanning the full lifecycle: patent search, document analysis, portfolio management, and IP intelligence.

A recurring theme was the importance of human-machine symbiosis. The objective, as panellists framed it, is not to replace the patent professional but to generate an “AI dividend”: achieving more with the same resources while continuously sharpening professional skills. 

The final AI-supported result remains an expression of the user's individual intellectual achievement. It is shaped by their understanding of the invention, their knowledge of patent law, and their verification of every output.

This is the principle at the core of how Solve Intelligence works. Every output, whether a drafted claim, a prosecution argument, or a prior art analysis, is presented to the user for review, refinement, and approval. The AI accelerates the work but the professional remains in control and owns the result. The AI dividend panellists described is real, but only when the tooling is designed to keep the professional in control at every step.

Technology and security considerations for AI systems

Our CEO, Chris Parsonson, joined this panel to discuss what it actually takes to deploy AI safely and effectively in IP workflows. The conversation built on a thread that ran through the entire forum. The value of an IP department's work lies in confidential, often unpublished, innovative ideas, and the tools used to process that information must be designed accordingly.

Confidentiality and data sovereignty

A central theme was the risk of feeding proprietary research, invention disclosures, or draft claims into public or third-party AI systems. Sensitive queries can risk premature disclosure long before a patent application is filed. General-purpose LLMs offer limited guarantees over how prompts and data are stored, reused, or exposed. 

This connected directly with Tobias Haar's spotlight session on sovereign AI: the ability of a nation or organisation to independently develop, control, and govern its own AI systems. It also linked to the “private vault” concept of keeping proprietary data within controlled infrastructure while still enabling AI-powered workflows. Data sovereignty, encryption, confidential computing, and role-based access controls are no longer optional. 

They are baseline requirements for any AI deployed in an IP context.

The limits of generic AI for patent work

The discussion also turned to the limits of generic AI for patent work. Three obstacles stand out:

• Hallucinations
• Shallow legal reasoning
• Lack of grounding in patent-specific data

Purpose-built infrastructure is the alternative. These are AI systems designed around patent drafting, prosecution, litigation, and search, with guardrails, auditability, and human-in-the-loop controls. That is what allows IP teams to move from experimentation to dependable, daily use.

Where Solve Intelligence fits in

The themes that ran through the forum (confidentiality, data control, the limits of generic AI, the rising bar for patent quality, and the need for human-in-the-loop workflows) are the problems Solve Intelligence was built to address.

Solve Intelligence is a purpose-built AI platform for patent professionals, spanning the entire patent lifecycle from drafting and prosecution to litigation. It's designed around the principles the forum's panellists described as non-negotiable: outputs grounded in patent-specific data, not generic LLM reasoning; client data that is never used to train models; and human-in-the-loop control at every step.

The forum made clear that the intersection of AI and IP will only grow more complex, and more consequential, from here. We're grateful to the organisers and fellow panellists for a thought-provoking day. We're committed to building the tools that help the profession meet what comes next.

See how Solve Intelligence handles the challenges your IP team is facing. 

Request a demo.

Frequently Asked Questions

Can AI be named as an inventor on a patent?

No. Following the DABUS decisions in the US, UK, Germany, and at the EPO, only natural persons can be named as inventors. AI is treated as a tool used by the human inventor.

What are the risks of using general-purpose LLMs for patent drafting?

Sensitive queries can risk premature disclosure before filing, and public LLMs offer limited guarantees over how prompts and data are stored, reused, or exposed. For patent work, that is rarely acceptable.

How does Solve Intelligence keep client data confidential?

Client data is never used to train models, infrastructure is built around the data sovereignty principles discussed at the forum, and every output is reviewed and approved by the professional before it leaves the platform.

For full details on Solve's security infrastructure, visit our Trust Center.

What is sovereign AI, and why does it matter for IP teams?

Sovereign AI is the ability to develop, control, and govern AI systems independently of third parties. For IP teams that handle unpublished inventions, that level of control is increasingly a baseline requirement, not a premium feature.