After attending the BioTechX conference in Basel last week, talking to people working across pharma, biotech, AI, software, laboratory automation and data management, my main takeaway was clear: the conversation has moved on from whether AI has a role in life sciences. The focus is now on how AI should be embedded in research and development, how its outputs can be trusted and where it creates a sustainable competitive advantage.
For companies developing AI-enabled technologies, this changes the IP question. It is no longer enough to ask whether an algorithm or model is patentable. The more useful question is where the real value lies across the whole technology architecture—and how that value should be protected and commercialised.
- The protectable value may lie beyond the AI model
One recurring theme was the move from conventional automation towards more agentic systems. Rather than simply following a predefined workflow, AI may help decide what should happen next, select tools, interpret intermediate results and adapt the process as it develops. In a biotech setting, a system might operate across target discovery, molecular modelling, data analysis and laboratory experimentation, while scientists retain oversight of the important decisions.
From an IP perspective, that makes it difficult, and often unhelpful, to identify the invention by looking at the AI component in isolation. The valuable technology may lie in the way the model interacts with biological or chemical data, other software, laboratory equipment and the experimental workflow. A new method of controlling a laboratory process, selecting between computational tools or translating biological data into a technical output may present a much stronger patent proposition than a generic claim to an AI algorithm.
This is particularly relevant in Europe, where patentability of computer-implemented inventions requires careful attention to technical contribution. The fact that software or AI is involved does not automatically exclude an invention from patent protection, but an interesting use of AI is not enough on its own. The technical problem, the particular implementation and the resulting technical effect all matter.
The practical lesson is to assess the system as a whole. For many AI-enabled life-sciences businesses, the strongest protection may be found at the points where software, data, biology and physical processes meet.
As a hypothetical example, consider a company that has trained a model to detect tumour cells in digitised pathology slides. A claim to “using a neural network to classify tissue images” may well face difficulties: the model may be generic, and classifying data on its own may be treated as abstract analysis. But suppose the system identifies regions of the slide where its confidence is low, directs the scanner to re-image only those regions at higher magnification or with a different focal plane, and incorporates the new images into its assessment. The invention is now a method of controlling an imaging process, with a measurable technical effect: higher-quality data where it matters most, without the time and storage cost of scanning the whole slide at maximum resolution. The AI is the same in both cases. What changes is where the claim locates the contribution.
- Trust, provenance and control can create the commercial moat
My BioTechX discussions also reinforced that AI in life sciences is not simply about producing an answer. Those answers must withstand scientific, regulatory and commercial scrutiny. If an apparently authoritative output can change materially with the framing of a prompt (or simply be wrong) then users need to understand where it came from, which data was used and how confidence in it was established.
Provenance, validation, traceability and explainability are therefore likely to become defining features of successful platforms. These features may themselves form part of the technical solution or valuable know-how. More broadly, proprietary datasets, domain-specific rules, validation mechanisms and workflow controls may create a more defensible position than access to a powerful underlying model that competitors can also obtain.
The protection strategy should reflect that mix. Patents may be appropriate for some technical implementations, while trade secrets may better protect processes that are difficult to reverse engineer. Database rights, copyright, contracts and practical controls over confidential know-how may also matter. At the same time, businesses need clarity over ownership and licensing where their products depend on third-party models, datasets, software or research inputs.
For founders and investors, the key due-diligence question is not simply how many patents the company owns. It is whether the company controls the technology and data on which its business depends—and whether it can continue to use them as the product and market develop.
- IP strategy needs to shape development, not follow it
A further theme was the growing use of AI within the IP workflow itself. Tools for patentability, competitive intelligence and freedom-to-operate analysis are increasingly being integrated into existing research environments. This creates an opportunity to identify relevant patent families, competing technologies and potential obstacles while there is still time to adjust the development path. It is worth noting that AI tools used with proprietary information should be carefully vetted to ensure that confidential inputs cannot be retained, disclosed or used inappropriately.
AI-powered workflows may be particularly valuable in biotech, where timelines are long and the cost of changing direction late can be significant. It also points to a broader change in the role of IP advisers. Rather than being asked only to protect a finished product, patent professionals can help identify where the technical differentiation lies, which developments are worth pursuing and how protection should align with commercial markets.
Patentability is only one side of that analysis. A strong portfolio has limited commercial value if third-party rights prevent the business from practising its technology. Freedom to operate, ownership, licensing and territorial filing strategy should therefore be considered alongside R&D and investment decisions, not only after them.
Looking at the complete architecture
What I found most striking at BioTechX was that AI no longer felt like a separate category of technology. It is becoming part of the infrastructure of life-sciences R&D, connected to biological data, molecular modelling, software pipelines, automation and physical instruments.
For companies working at the intersection of AI, biotech, pharma, diagnostics or laboratory automation, the challenge is not simply to ask whether the AI can be patented. It is to identify where the genuine competitive advantage lies across the complete architecture, choose the right combination of rights to protect it and establish freedom to commercialise it.
Those decisions are most valuable when they influence development, rather than document it after the event. An early IP review can help a business focus investment on the most defensible aspects of its technology, address ownership and freedom-to-operate risks, and build a stronger story for partners and investors.
If you are working in this space, I would be happy to hear more about what you are doing and chat through your IP position – whether that is to identify innovations, clarify ownership and freedom-to-operate risks, align IP strategy with investment and market priorities, or any other IP question you may have.
This article was prepared by HGF Partner and Patent Attorney Janine Swarbrick.