Divided Infringement: When No Single Party Performs Every Step

Divided infringement arises when the steps of a method claim are performed by more than one party, such as a customer and a service provider, and no single party performs them all. Because direct infringement of a method generally requires that every step be performed by, or attributable to, one entity, a claim drafted this way may be infringed by nobody, even when the whole method is clearly being used. Careful drafting can reduce the problem by framing claims around steps performed by a single actor, where the invention and the original disclosure support doing so.

This issue comes up constantly in software, networked, and connected-device inventions, where operations are naturally split between devices and parties.

Why it happens

Engineers describe systems the way they work, with users clicking, devices sending, servers processing, and third-party services responding. A method claim that follows that description closely might read like this:

A method comprising:
a user selecting an item on a mobile device;
transmitting the selection to a server;
the server generating a recommendation based on the selection; and
displaying the recommendation on the mobile device.

That claim describes the invention accurately. The problem is who performs each step. The user selects the item, the user’s device transmits and displays, and the provider’s server generates the recommendation. If the user and the provider are separate parties, neither one performs all of the steps.

What the courts have said

Two decisions in the same long-running dispute set the current framework.

In Limelight Networks, Inc. v. Akamai Technologies, Inc., 572 U.S. 915 (2014), the Supreme Court held that a party cannot be liable for inducing infringement of a method claim under 35 U.S.C. 271(b) unless someone has directly infringed under 35 U.S.C. 271(a). So inducement cannot fill the gap when no single party directly infringes.

On remand, in Akamai Technologies, Inc. v. Limelight Networks, Inc., 797 F.3d 1020 (Fed. Cir. 2015) (en banc), the Federal Circuit explained when the steps performed by others can be attributed to a single entity. That happens when the entity directs or controls the others’ performance, such as through an agency relationship or a contract, or when the parties form a joint enterprise. The court also held that attribution can occur when an alleged infringer conditions participation in an activity, or receipt of a benefit, on performing a step, and establishes the manner or timing of that performance.

That framework helps in some situations, but whether it applies depends on the facts of each case. A claim whose infringement depends on proving direction or control is generally harder to enforce than one a single party clearly performs.

System claims are not a complete answer

System claims raise related questions. In Centillion Data Systems, LLC v. Qwest Communications International, Inc., 631 F.3d 1279 (Fed. Cir. 2011), the Federal Circuit held that a party uses a system when it puts the system as a whole into service, controlling it and obtaining a benefit from it, which can allow enforcement even when different parties own different components. Making a system generally requires one party to assemble all of its elements, and other acts, such as selling, raise their own questions. A system that spans a customer’s device and a provider’s servers can be enforceable against one party in one situation and not in another.

Drafting single-actor claims

A useful drafting approach is to frame each method claim around operations performed by a single actor, where the invention and the original disclosure support doing so, describing what that actor does and what it receives from others.

Server-side claims. The recommendation example above can be reframed around operations performed on the provider’s servers:

A method comprising:
receiving, by one or more servers from a client device, data indicating a selection of an item;
generating, by the one or more servers, a recommendation based on the selection; and
transmitting, by the one or more servers, the recommendation to the client device for display.

The user’s action becomes data the servers receive, and display becomes something the servers enable. Reciting “one or more servers” rather than “a server” reflects the Federal Circuit’s reasoning in Salazar v. AT&T Mobility LLC, No. 2021-2320 (Fed. Cir. Apr. 5, 2023), where a claim reciting “a microprocessor” that performed several functions was read to require at least one microprocessor capable of performing all of them. Real services spread work across many machines. Framed this way, the claim does not separately require steps performed by a user or a client device, which can reduce the need to establish attribution of steps performed by multiple parties. Whether a particular entity directly infringes still depends on the claim language and the facts of the accused implementation, such as how the service is actually architected and who operates each part of it.

Client-side claims. A parallel claim can cover what the client device does: detecting the selection, sending it, receiving the recommendation, and displaying it. Note the distinction between performing a method and supplying software capable of performing it. The device or user that runs the steps performs the method; a company that distributes client software may face questions of indirect infringement rather than direct performance, which brings in the knowledge requirements for inducement and contributory infringement.

Receiving instead of requiring. Where another party’s action matters, recite receiving its result rather than performing it. “Receiving a signal from a sensor” is performed by the receiving device. “A sensor generating a signal” is performed by the sensor, which may belong to someone else.

Avoid steps performed by users. Claim steps such as “a user entering” or “the user approving” introduce a second actor. Where user input matters, recite receiving or detecting it.

Pair methods with other claim types. Computer-readable medium claims and apparatus claims covering the same operations do not raise the same step-performance issue, because they are directed to the stored instructions or the device itself, although they have their own infringement questions.

Reviewing existing claims

If an issued patent’s claims require multiple actors, a pending continuation offers a chance to pursue single-actor claims, as long as the original specification supports them. Specifications that describe each component’s operations separately, from both the client and server perspectives, make that much easier. That is one more reason to describe each actor’s role in detail when drafting the original application.

A practical takeaway

For each method claim in a draft, ask who performs every step. If the answer involves more than one party, ask your patent attorney or agent whether claims framed around each party’s operations would be supported and useful. That review takes little time before filing and can avoid a problem that is difficult to address afterward.