02/07/2026
Swarm intelligence is an emerging technology characterised by decentralised control and biomimetic mathematics, where emergent behaviour arises from localised rules, providing a macro-scale intelligence similar to a hive mind in an insect colony. But patenting swarm-intelligence inventions at the EPO is difficult because their nature is contrary to the expectations of patent systems.
Swarm intelligence is used for optimisation algorithms (such as ant colony optimisation and particle swarm optimisation) to solve problems that are too complex for brute force computing. It is used for swarm robotics where, in stead of building one massive and expensive robot, engineers build a swarm of cheap micro-robots that communicate with each other. It is used for telecommunications and networking, where swarm logic is used to manage network traffic. If a link or node goes down, digital ‘ants’ in the form of software agents reroute data packets through the network to prevent outages. And we are seeing it used in defence and aerospace, where militaries are actively developing autonomous drone swarms, which have no central command to jam or hack or destroy, and so a swarm of small drones can autonomously navigate, monitor an area and react to threats as a single, fluid entity.
But while we are seeing this architecture having a great effect in these fields, swarm-intelligence inventions face obstacles when seeking patent protection. The patent systems were historically designed for deterministic, self-contained physical machines, whereas swarms are distributed, non-deterministic and often heavily abstract. This mismatch makes it difficult to protect swarm intelligence inventions through patents. But it’s not impossible. Applicants can follow deliberate drafting strategies to increase their chances of getting a patent for their swarm intelligence invention.
Divided infringement and single-node claiming
A swarm is decentralised by its nature. When no single party makes, sells or operates a whole swarm, a patent directed to a swarm is difficult to enforce. The bar for showing contributory infringement or joint tortfeasorship is high in European courts.
To reduce this enforcement risk, we recommend applicants focus protection on a single node. The patent application should be drafted from the perspective of an individual hardware unit. A claim should not be “A swarm comprising: …” but instead perhaps “An autonomous node configured to …”. This way a patent owner can target any of the multiple infringing parties directly. We have previously written about drafting strategies for Internet of Things inventions and the same principles apply to a swarm intelligence
The emergence problem: sufficiency of disclosure (Article 83 EPC) and inventive step (Article 56 EPC)
The problem of emergent behaviour in sufficiency of disclosure is swarm-specific. A swarm-intelligence solution relies on emergent behaviour to solve a problem, without a central plan made in advance. We can’t predict the exact steps a swarm will take. The swarm behaviour can be something of a black box, hopefully providing a desired end result. This is similar to the sufficiency and inventive step issue in AI inventions and arises because a neural network is also a black box where it’s not prima facie clear that an arbitrary neural network will solve a desired problem. At the EPO, claims must comply with sufficiency and inventive step requirements across their whole scope. This means that the claim scope should conform with what is actually taught in the application such that an advantage or effect is achieved across the whole scope of the claim.
The solution we recommend is to tightly define the microscale behaviour to mathematically guarantee the macro outcome. We must shift the focus from an unpredictable global outcome to predictable local rules. For example, we can define deterministic rules or threshold values that an individual agent or node uses. We can specify in the claim that a drone executes an avoidance movement if a nearest neighbour is closer than 5 metres. In addition, we can specify fallback safety protocols that guarantee a technical effect even if the emergent behaviour fails or gets stuck in a local minimum.
Mathematical method exclusion (Article 52 EPC)
The algorithms governing a swarm’s behaviour and that give rise to a swarm intelligence are mathematical models inspired by nature. Mathematical methods may provide advantages but they are not protectable by themselves. Following the EPO’s well-established COMVIK approach and decision G1/19 of the Boards of Appeal, mathematical methods by themselves are excluded from patentability and claim features that are merely mathematical methods can be ignored when assessing inventive step.
To overcome this, the abstract mathematics must be applied to a specific real-world technical effect. For robot or drone technology, this can be done straightforwardly by controlling a rotor speed based on the output of an algorithm. The invention is then not merely a mathematical method but is a method for controlling the movement of a machine. Even in cases of particle swarm optimisation or ant colony optimisation, if the output is used to control a telecommunications network then the output is technical and the mathematical features of the claim may be included in the inventive step assessment. Where we see more difficulty is where the mathematical methods are used as a simulation or for processing arbitrary data, where patentability will be on more of a case-by-case basis, and may require us looking at how the mathematical features take into account the underlying hardware upon which they operate.
Conclusion
Swarm intelligence inventions are not easy to patent. They require careful drafting to survive scrutiny from patent offices such as the EPO. They face obstacles as discussed above but these obstacles may be overcome by careful drafting and strategic amendments and arguments during examination, which is where Reddie & Grose can help. It is not necessary to include all of the elements mentioned above in the claims on filing, particularly if an application is initially drafted for the USPTO, but it is important that these elements are included in the specification to support future amendments and arguments in Europe and elsewhere.
If you are managing a portfolio involving swarm intelligence or drone technology, or have a particular case to discuss, get in touch with the specialist attorneys at Reddie & Grose.
This content is for general information only. Its content is not a statement of the law on any subject and does not constitute advice. Please contact Reddie & Grose LLP for advice before taking any action in reliance on it.
