Liang-Zhi International IP Center

The recent dismissal of Tempus AI v. Guardant Health in the Northern District of California (Jan 2026) is a wake-up call for every GC and IP Strategist in the Bio-IT space.

Judge Trina Thompson just threw out Tempus AI’s infringement suit, finding their AI-driven diagnostic patents ineligible under 35 U.S.C. § 101.

The Hard Truth: Simply “applying AI” to complex biological data doesn’t make it patent-eligible. The court viewed Tempus’s claims—which focused on analyzing genetic data to characterize tumors—as “abstract ideas” and “mental processes.”

The Litigation Fallout: This isn’t just a loss of a lawsuit; it’s a loss of leverage. Tempus filed this as a counter-offensive to Guardant’s Delaware suit. By losing these patents at the 12(b)(6) stage, Tempus has lost its primary “bargaining chip” for settlement, leaving them exposed in the ongoing battle over Guardant’s liquid biopsy portfolio.

The Prosecution Lesson: If your claims describe what the AI does (e.g., “identifying a correlation”) rather than how the AI architecture itself is technically improved or how it physically transforms a process, you are at high risk of a § 101 “Alice” rejection—or worse, an early dismissal in court.

Questions every IP team should be asking right now:

1️⃣ Do our claims recite a specific technical improvement to a machine-learning system—or just a generic model applied to a new dataset?
2️⃣ Are we effectively claiming a natural law or correlation, or the technical tool used to uncover it?
3️⃣ Is the computation meaningfully tied to a physical step (e.g., a novel sequencing, sample prep, or signal-generation process)?In the age of AI, “black box” patents are a house of cards.

Durable portfolios are built on technical architecture, system-level improvements, and physical integration—not just computational outcomes.

Case Name: Tempus AI, Inc. v. Guardant Health, Inc.

Case Number: 3:25-cv-06622 (N.D. Cal.)