As the Patent Office embraces a “commitment to expansive eligibility” and Congress considers the Patent Eligibility Restoration Act (PERA), which would narrow courts’ ability to invalidate claims under Section 101, a central question is whether broader patent eligibility will unleash patents that recite desired outcomes to be achieved with artificial intelligence without disclosing technology improvements to achieve them—or that merely instruct practitioners to “do it with AI.” The Federal Circuit’s recent decision in Brita LP v. ITC offers a practical answer: rigorous enforcement of Section 112’s written description requirement can check result‑oriented, functionally claimed AI patents that fail the patent system’s disclosure bargain.

Brita involved a broad, functionally framed claim—any gravity-fed water filter with a filter media having specific properties that achieves a performance metric—where the specification taught only one working technology. The Federal Circuit affirmed invalidity on Section 112 written description (and enablement) grounds and, in doing so, articulated principles that map onto attempts to patent desired results or outcomes achieved with AI. 

First, the court emphasized that functional results cannot substitute for technological possession. The panel explained: “It is not enough for the patent to simply mention a desired outcome …,” and that “the written description requirement demands that the specification provide more detail than merely listing examples.” 

Second, the court made clear that broad, functional genus claims require written description support commensurate with their scope. Citing Ariad, the court reiterated that for a functional genus, the patentee must show either “a representative number of species falling within the scope of the genus or structural features common to the members of the genus so that one of skill in the art can visualize or recognize the members of the genus,” and that the problem “is especially acute with genus claims that use functional language.” It held that the patent at issue “does not identify a representative number of species or common structural features” beyond the one working species. 

Third, the court underscored the quid‑pro‑quo logic behind written description as the backstop against claiming results. It reaffirmed that the specification must show the inventor “had possession of the claimed subject matter as of the filing date” and rejected gap‑filling by artisans: “it is the specification, not the knowledge of one skilled in the art, that must supply” what is claimed; a “mere wish or plan” does not suffice. And the patentee must “demonstrate [he] has invented species sufficient to support a claim to the functionally‑defined genus.” 

Taken together, those passages show how a reinvigorated Section 112 written description doctrine can address the AI‑specific concern that an eligibility reset will permit patents claiming AI outcomes rather than AI technology. Even if more AI claims clear Section 101, Brita teaches that functionally defined claims must still be supported by a specification that demonstrates possession of the claimed genus—through representative AI “species” or disclosed common structural features—rather than aspirational statements. 

Applying the lesson to artificial intelligence where functional, result‑oriented claiming could be broadly preemptive without disclosing any actual technological advance, Section 112’s written description requirement should demand concrete technical disclosure commensurate with AI claim breadth. A patent that discloses only a desire to achieve a result “with AI” (e.g., by invoking a generic neural network, machine learning, or artificial intelligence) runs afoul of Brita's principles:  “[i]t is not enough for the patent to simply mention a desired outcome” and for broad, functional claims the patentee must show he “has invented species sufficient to support a claim to the functionally‑defined genus.” Those statements translate to AI: reciting objectives and adding “use AI” is a “mere wish or plan,” not a demonstration of possession of a functionally defined AI genus. 

Brita's core message is that Section 112’s written description requirement is not a formality. It keeps patents anchored to real technology. Independent of the state of Section 101, courts can and should apply Section 112’s written description requirement with rigor to ensure that AI patents reward disclosed technological advances, not aspirations. That approach preserves incentives to advance AI technology while protecting the public’s right to practice what has not been taught.