Key Considerations for Patenting Software at the EPO
When seeking patent protection for software-related inventions at the European Patent Office (EPO), applicants must navigate specific requirements that go beyond standard computer implementation.
Understanding these principles helps ensure that claims are drafted effectively and that the invention demonstrates the necessary technical contribution. Below are some of the most important aspects to keep in mind:
At the EPO, software patents succeed only when they demonstrate a genuine technical effect beyond standard computer use.
1. Eligibility/Patentable Subject Matter Requirements
The EPO requires software inventions to demonstrate a “technical effect” beyond normal computer operations (e.g., improved data processing with resource efficiency, or hardware control).
Example: A software algorithm optimizing battery life would satisfy the EPO’s technical effect requirement.
2. Claiming Signals as well as Non-Transitory Media
The EPO allows claims directed to signals or electromagnetic waves (e.g., encoded data in
transmission) if the data carried can achieve a technical effect – explicit disclosure of both transitory and non-transitory media should be provided in the priority filing.
Example: A computer-readable medium storing instructions that, when executed by a computer, cause
the computer to perform a method comprising steps A, B, and C.
3. Disclosure Requirements for Priority
The EPO requires all integers of the claimed invention to be disclosed in combination in the priority application for valid priority claims.
Example: Adding a new algorithm step during EPO prosecution could invalidate priority if not originally
disclosed in combination with other features of the proposed claim.
4. Inventive Step through PSA
The EPO uses a problem-solution approach (PSA):
Identify the closest prior art and the difference compared to the claimed invention.
Define the objective technical problem solved by the claimed difference.
Assess whether the solution is obvious to the skilled technical person (e.g. using common general knowledge (cgk) or a second reference).
Example: A software feature improving data compression in a specific application which provides beneficial storage, transmission or speed improvements have been succesfully argued.
5. Enablement or Sufficiency of Disclosure
The EPO requires sufficient disclosure for reproducibility of the invention by a skilled person “without undue burden”. The claimed invention must be reproducible across its full scope using only the application’s disclosure and common general knowledge in the field at the effective filing date. Importantly, the EPO does not permit general reliance on “incorporation by reference”— all technical detail necessary for implementation must be present in the application as filed.
6. Sufficiency for ML innovation
In the case of machine learning (ML), this typically requires a higher degree of specificity than might be expected in other jurisdictions. The description should include the following:
Data management protocols – how the training data was selected, filtered, pre-processed, and structured.
Dataset size and quality – sufficient information to understand the scale, representativeness, and statistical value of the data.
Learning architecture and algorithms – the type and configuration of the model, such as neural network type or optimization techniques.
Hardware requirements – where pertinent, the computing environment needed to implement or train the model.
Training methodologies – including training objectives, loss functions, regularisation, and convergence strategies.
Refinement or tuning procedures – how the model is iteratively improved or validated.
If such information is absent or too vague—particularly where the claimed invention relies on non-standard approaches or where the performance depends on unknown variables—the EPO may raise a lack of sufficiency objection, even if the invention would likely pass enablement or written description tests elsewhere.
Ask us for our improved IDS format for ML developers – it really gets improved innovation disclosure upfront.
Strategic Advice
When drafting EPO filings (or PCT applications intended for Europe), ensure that the disclosure is self-contained and technically detailed.
Avoid merely generic descriptions of software and ML processes and assume that reviewers will require a clear path to reproduction. Include examples, schematic workflows, and fallback positions (e.g., claims with more defined model architectures) to withstand scrutiny.