In Short

The Situation: Artificial intelligence ("AI") and big data ("BD") innovations are driving forces of the current technological revolution, dramatically changing the way we search for information, communicate, operate devices, manufacture things, and solve problems. Companies have traditionally turned to patents as the strongest mechanism to protect their innovation and secure investment.

The Issue: The U.S. Supreme Court's decision in Nautilus, Inc. v. Biosig Instruments Inc. (Sup. Ct. 2014) has made it easier to invalidate patent claims on grounds of indefiniteness, and the Federal Circuit's decision in Williamson v. Citrix Online (Fed. Cir. 2015) has made it easier for claims to have a "means-plus-function" interpretation even where unintentional. Patent coverage for AI- and BD-related innovations relies heavily on functional claiming, and successful functional claim drafting requires attention and strategies for navigating the potential pitfalls.

Looking Ahead: While patents are not necessarily the only, or the best, protection in a given instance for AI and BD innovation, where patent protection is sought, practitioners need to pay attention to issues of functional claiming as well as issues of subject matter eligibility under Alice in order to execute successful patent coverage for AI and BD innovation.

Companies are increasingly investing in AI innovations to reduce costs and increase accuracy for tasks and analyses that might otherwise require human intelligence and judgment. These AI innovations are becoming particularly important as computers, sensors, and other devices generate massive amounts of data (i.e., big data) that needs to be characterized for or otherwise consumed by various software applications and computing devices.

A byproduct of the increased investment in AI has been a significant uptick in the number of worldwide patent filings directed to AI and BD. A previous Commentary presented an overview of trade secret and copyright protections for AI in relation to patent protection, and a related Commentary presented some specific patenting considerations and strategies for navigating subject matter eligibility under Alice Corp. v. CLS Bank Int'l (2014). Here we offer some additional considerations for obtaining patent protection on such innovations relating to functional claiming.

Functional Claiming

In addition to ensuring that the claims are directed to patent-eligible subject matter, extra care needs to be taken to ensure that the scope and coverage of claims for AI and BD innovations are precise. Attention to language precision is particularly important because functional claiming is a practice typically utilized in computer-implemented inventions, as it affords the flexibility of describing a desired result without tying it to one of many specific software solutions. But this drafting convenience may come at a steep price in either prosecution or litigation if the drafter does not account for several potential risks.


Previously, claims were invalid as indefinite if "not amenable to construction" and "insolubly ambiguous"—a very difficult standard to meet. Now, patent claims must inform those skilled in the art about the scope with reasonable certainty. (Nautilus, Inc. v. Biosig Instruments Inc. (Sup. Ct. 2014)). Functional claiming that may be attacked as indefinite under Nautilus includes terms such as "adapted to," "operable to," "configured to," etc., especially if specification provides little detail. Traditionally, such terms are read in light of the specification to convey the meaning of design or configured to accomplish the specified objective. But potential problems can arise if the specification is written such that these terms could be construed as "capable of," without disclosing actual implementation details.

Unintended "Means-Plus-Function" Interpretation

Section 35 U.S.C. 112 provides that an element in a claim "may be expressed as a means" for performing a specified function without further recitations, but it also specifies that such claim element "shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof." Until recently, there was a strong presumption that if the patent drafter used the words "means for" performing a function, it triggered Section 112; if these words were lacking in the claim, Section 112 generally was not deemed triggered. Now, whether a claim element should be construed under Section 112 depends on "whether the words of a claim are understood by persons of ordinary skill in the art to have a sufficiently definite meaning as the name for structure." (Williamson v. Citrix Online (Fed. Cir. 2015)).

Under Williamson it is much easier for certain language to invoke Section 112 interpretation. Consider, e.g., "module for," "component for," "logic for," etc. If such elements are construed (unintended) under Section 112, one may end up with unexpected claim scope closely tied to the specification disclosure or lose the claim as indefinite if the specification contains no structure corresponding to the claimed function. In addition, for a software function under a Section 112 interpretation, the corresponding "structure" is the algorithm disclosed as performing that function in the specification. But if no such algorithm is actually disclosed in the description (including figures), the limitation may be found indefinite. This can be esspecially problematic for simple well-known functions where the drafter was not contemplating a Section 112 interpretation. It does not help if an algorithm for the function would have been readily known to a person of ordinary skill in the art—if the algorithm as the corresponding "structure" for a claimed function is not disclosed, the limitation may be found indefinite if construed as a means element.


Functional claiming may also involve a risk that a claim element may be interpreted too broadly. Issues of overbreadth can arise during prosecution and may be addressed with amendments and arguments regarding claim scope to overcome the prior art being addressed at that time. Overbreadth presents even more of a concern in litigation, however, where granted claims may be interpreted to have a breadth that covers unintended prior art, risking invalidity. To mitigate, claim drafters can include a spectrum of dependent claims or other claims of varying and narrower scope to provide backup positions, particularly for the functionally claimed elements.


AI and BD inventions can be effectively and concurrently protected via patents, trade secrets, and copyrights, as well as by means of contractual arrangements. Notwithstanding such variety of protection mechanisms, patents remain an essential part of an overall IP strategy for innovation-oriented companies. Given uncertainties over patent eligibility and validity, careful attention must be heeded to optimally capture the value of AI and BD innovations for both defensive and offensive purposes.

Three Key Takeaways

  1. Use care with generalized functional claim terminology and prepare the patent specification to explain functional limitations and provide at least some examples of implementation.
  2. The specification should contain sufficient disclosure for each claim limitation that may be deemed "functional," whether you intend it as such or not (e.g., how variables for models are extracted, how models are trained, how models utilize data after training, how outputs of models are used, etc.).
  3. Use care with claim terminology and consider how the scope of a term may be interpreted in future litigation. Utilize a claiming strategy in which there are dependent claims having clear, meaningful, and narrower scope as a backup against claims being interpreted later in an overly broad manner and invalidated.