The Supreme Court Reins In Secondary Liability

The defining trend of the term is a Court unwilling to let patent and copyright owners reach past the direct infringer to the platforms and competitors around them.

On March 25, 2026, in Cox Communications, Inc. v. Sony Music Entertainment, the Court held that internet service providers cannot be held indirectly liable for their customers’ copyright infringement merely because they knew infringement was occurring but failed to stop it. Notably, the majority opinion leaned heavily on patent-law doctrine to frame the new two-pathway test for contributory copyright liability — inducement, or a service specifically tailored to enable infringement — stating that these theories “track patent law” and walking through the parallels in detail.

That cross-pollination proved prophetic. Ten weeks later, on June 4, 2026, the Court ruled unanimously the other direction on the patent side in Hikma Pharmaceuticals USA Inc. v. Amarin Pharma, Inc., reaffirming existing pleading standards while reversing the result below — another loss for a rights holder pressing a secondary or induced-infringement theory. Commentators have read the two decisions together as a single signal from One First Street: the Court has now rejected attempts by both patent and copyright holders, in the same term, to impose secondary liability on companies accused of inducing infringement by others.

For practitioners, the takeaway is structural, not just doctrinal. Enforcement strategies built around “the platform should have known” or “the generic competitor should have known” now face a considerably higher bar in federal court, and complaints should be re-drafted with that reality in mind before filing.

AI Training Data Heads Toward a Reckoning

If secondary liability was the term’s quiet story, AI training data was the loud one. Generative AI copyright litigation has moved from motions to dismiss into the summary-judgment stage that will actually decide the fair-use question at the heart of the industry.

The clearest signal so far came from the music industry. Two early California rulings found that training AI on copyrighted books can constitute fair use, but that storing pirated copies does not — and one such ruling produced a settlement worth roughly $1.5 billion, translating to an estimated payout near $3,000 per work. That result set the template other plaintiffs are now racing to either extend or distinguish.

The next major test lands imminently: a summary-judgment hearing in Sony Music’s suit against the AI startups Suno and Udio, expected in the District of Massachusetts before Chief Judge F. Dennis Saylor IV, is being watched as a bellwether for whether unlicensed AI training on copyrighted works is lawful across text, images, code, and video — not just music. Suno is leaning on the book-training precedent, while the recording industry counters that generating competing music fails the fourth fair-use factor governing market harm, because analysis and generation are not the same thing. Adding to the pressure, a federal court in May allowed key claims under the Digital Millennium Copyright Act to proceed against Udio regardless of how the fair-use question is ultimately resolved, and independent-artist plaintiffs have now brought in major class-action counsel to press allegations that stream-ripping software was used to build the training sets in the first place.

Litigation is also broadening well beyond music. In mid-July, a group of publishers and authors — including Hachette, Cengage, Elsevier, and author Scott Turow — sued Google over the training data behind its Gemini models, alleging that copyright management information was stripped from the works to conceal their use. On the legal-technology front, the Federal Circuit is separately weighing an appeal in Thomson Reuters v. Ross Intelligence, where a trial court had already found that training a competing legal-research tool on Westlaw headnotes was not protected by fair use — a ruling that, if affirmed, would sharpen the line between “learning” and “competing.”

The throughline across every one of these cases is the same four-factor test under 17 U.S.C. § 107, with courts increasingly turning on one question: does the resulting AI system substitute for the market the copyright owner controls, or merely learn from it? Expect that market-substitution inquiry, not transformativeness in the abstract, to decide most of what’s left on the docket in the back half of 2026.

No Room for the Machine as Author or Inventor

While training-data litigation plays out, the Court has been unambiguous about keeping AI systems themselves out of the ownership column. On March 2, 2026, the Supreme Court denied certiorari in Thaler v. Perlmutter, ending Dr. Stephen Thaler’s yearslong effort to secure copyright protection for artwork he says was created autonomously by his AI system, DABUS. The denial leaves undisturbed the rule that AI cannot be listed as an “author” for copyright purposes, and the same human-inventorship principle continues to govern on the patent side at the USPTO. The practical guidance for businesses building IP portfolios around AI-assisted work has not changed: document the human contribution behind every invention or creative output, because recordkeeping — not the sophistication of the tool — is what will preserve eligibility for protection.

Patent Eligibility Stays Unsettled, and the Federal Circuit Keeps Score

Section 101 patent eligibility remains the least settled corner of IP law. The Federal Circuit’s Recentive decision reinforced that applying generic AI or software techniques to a new data environment, dressed in results-oriented claim language, remains ineligible absent a genuine technological improvement to the model or computer itself, even as the USPTO’s own Desjardins framework nudges examiners toward treating Section 101 as a gatekeeper and resolving breadth and inventiveness questions instead under Sections 102, 103, and 112. The result is a widening gap between a friendlier allowance path at the agency and continued skepticism from the courts — a divergence worth watching as Congress again floats the Patent Eligibility Restoration Act and companion bills aimed at patent reform.

On the litigation side, the Federal Circuit has kept up a brisk pace on pharmaceutical patents specifically. On July 1, 2026, in Otsuka America Pharmaceutical, Inc. v. Hetero Labs Limited, the court affirmed a district court’s claim construction and preliminary injunction, rejecting a generic manufacturer’s proposed construction because it would have produced the “highly improbable” result of excluding the patentee’s own Orange Book–listed drug from the claims. A week earlier, in Enanta Pharmaceuticals, Inc. v. Pfizer Inc. (June 23, 2026), the court affirmed summary judgment invalidating a patent for lack of adequate written-description support in its priority provisional application, clarifying that the test is whether the provisional demonstrated possession of the later-claimed subject matter — not whether a discrepancy could be waved off as a typo. Both cases are useful reminders that drafting discipline at the provisional and claim-construction stage still decides pharmaceutical patent fights years later.

On the procedural side, the USPTO revised its practice, effective June 24, 2026, for petitions based on unintentional delay, changing when the higher petition fee applies — a small but real change for prosecution counsel managing lapsed deadlines.

Agency Housekeeping Worth Tracking

Away from the courts, both IP agencies are moving on their own initiatives. On June 24, 2026, the American Intellectual Property Law Association submitted comments on the U.S. Copyright Office’s Notice of Inquiry into alternative copyright registration fee structures, part of a broader agency review of how registration is priced and administered. The Copyright Office and USPTO leadership have also been active on the public-engagement front, participating alongside congressional IP subcommittee members in World IP Day programming — a signal that both agencies are positioning themselves for a legislative push on AI and patent reform later this year.