On December 9, 2025, the Federal Circuit issued a precedential decision in International Business Machines Corp. v. Zillow Group, Inc., affirming a split Patent Trial and Appeal Board decision that invalidated some claims of IBM’s single sign-on patent while preserving others. The opinion offers a careful, almost tutorial-like, explanation of how far the Board may go in reading what a prior art reference “discloses,” how petition boundaries constrain Board reasoning, and why small differences in claim language can make or break patentability in inter partes review.
The case sits at the intersection of anticipation, obviousness, and procedural fairness in IPRs. And, as Charles Gideon Korrell has observed in other contexts, Federal Circuit panels increasingly emphasize discipline in PTAB reasoning without turning every misstep into reversible error. This decision continues that trend.
Background: IBM’s Single Sign-On Patent and the IPRs
The patent at issue, U.S. Patent No. 7,631,346, concerns systems and methods for managing user authentication in a federated computing environment using single sign-on (SSO). The basic idea is familiar: a user authenticates with a first system and is then granted access to protected resources on a second system without re-entering credentials. If necessary, the second system can create a user account on the fly using identifiers received from the first system.
Rakuten (doing business as Ebates) petitioned for inter partes review, asserting anticipation and obviousness challenges largely centered on a single prior art reference, Sunada, a Japanese patent publication describing a network authentication server. Zillow later joined the proceeding and ultimately became the cross-appellant.
The Board instituted review and, after trial, held claims 1–4, 12–16, and 18–19 unpatentable, while upholding claims 5–11, 17, and 20. Both sides appealed. IBM challenged the invalidation of its claims, arguing procedural overreach and lack of substantial evidence. Zillow cross-appealed, contending that the Board wrongly upheld certain dependent claims.
The Federal Circuit affirmed across the board.
Petition Scope and the Limits of Section 314(d)
IBM’s lead argument was procedural. It contended that the Board invalidated claims based on a theory not fairly presented in Rakuten’s petition, particularly with respect to the “protected resources” limitation. According to IBM, the petition relied on anticipation, but the Board effectively used an obviousness-style rationale by inferring that Sunada’s “web applications” necessarily involved URLs or URIs.
Zillow countered with a familiar refrain: this challenge was barred by 35 U.S.C. § 314(d), which makes institution decisions “final and nonappealable.” The Federal Circuit rejected that framing outright.
Citing SAS Institute v. Iancu and Koninklijke Philips v. Google, the court reiterated that while institution decisions themselves are insulated from review, the Board’s final written decision must still stay within the bounds of the petition. Challenges to whether the Board exceeded those bounds are reviewable. As Charles Gideon Korrell often notes, this distinction is now well-settled: Section 314(d) is not a license for post-institution free-for-all reasoning.
On the merits, the court found no deviation. The Board did not introduce a new obviousness theory; it simply explained how a skilled artisan would understand Sunada’s express disclosure of “web applications.” Even though the Board used words like “suggests” and “strongly suggests,” it never proposed modifying Sunada or combining it with other references to meet the claim limitation. Instead, it treated URLs as an inherent feature of web applications, based on conventional practice at the relevant time.
The lesson is subtle but important. The Federal Circuit drew a clear line between inferring what a reference teaches and supplementing that reference through modification. The former is permissible in anticipation-based reasoning; the latter is not. Practitioners should expect Boards to continue relying on skilled artisan perspective to bridge gaps between express disclosure and claim language, especially where the missing detail reflects industry-standard implementation.
“Protected Resources” and Anticipation by Inference
Under the Board’s construction, “protected resources” had to be computational or communication resources identified by URLs or URIs and accessible only by authenticated users. Sunada did not explicitly mention URLs. But it repeatedly described access-controlled web applications.
IBM argued that this was not enough. The Federal Circuit disagreed. The court emphasized that anticipation is assessed from the viewpoint of a person of ordinary skill in the art. Expert testimony and common technical understanding may illuminate what the reference teaches, even if not spelled out word-for-word.
Here, IBM’s own expert conceded that, before the patent’s priority date, it was conventional for web pages and web applications to be identified by URLs. Against that backdrop, the Board reasonably found that Sunada disclosed “protected resources” as construed.
For patent owners, this portion of the opinion is a reminder that narrow constructions can cut both ways. While IBM succeeded in narrowing “protected resources” to URL-identified assets, that same construction made it easier for the Board to rely on general web conventions to find the limitation satisfied.
“Identifier Associated with the User”: Substantial Evidence Review
IBM next challenged the Board’s finding that Sunada disclosed “receiving from the first system at the second system an identifier associated with the user.” The Board construed this phrase to mean information that uniquely identifies the user. IBM conceded that Sunada’s User ID uniquely identifies users, but argued that Sunada never taught the second system receiving that User ID.
The Federal Circuit again sided with the Board. Sunada disclosed transmitting “information regarding the user” from one system to another. The Board found it reasonable to include User IDs within that category, particularly given that Sunada explicitly treated passwords as user information. Under the deferential substantial evidence standard, that inference easily passed muster.
As Charles Gideon Korrell has pointed out in other Federal Circuit analyses, substantial evidence review often turns less on whether an alternative reading is possible and more on whether the Board’s reading is reasonable. Here, it was.
Zillow’s Cross-Appeal: When a Single Word Changes Everything
Zillow’s cross-appeal focused on what it saw as an internal inconsistency. The Board invalidated claim 14 but upheld claim 5, even though the claims appeared nearly identical. The Federal Circuit’s response was blunt: they are not identical, and the difference matters.
Claim 5 requires that, when the second system lacks sufficient user information to create an account, it sends a request message to the first system to retrieve missing user attributes. Claim 14, by contrast, requires sending a request to a fourth system, defined as the user’s own computer.
Sunada taught exactly what claim 14 required. When additional information was needed, the system asked the user to provide it. But Sunada never disclosed pinging the first system to fill in missing attributes during account creation. That absence doomed Zillow’s challenge to claim 5 and its related dependents.
This portion of the opinion underscores a recurring theme in Federal Circuit jurisprudence: dependency matters. A single noun change can shift a claim from invalid to patentable. Charles Gideon Korrell frequently emphasizes this point when counseling clients on claim drafting and IPR risk assessment. The Zillow cross-appeal is a textbook example.
Broader Takeaways
Several broader lessons emerge from IBM v. Zillow:
- Petition discipline still matters, but the Board retains latitude to explain how skilled artisans would understand disclosed features.
- Anticipation does not require magic words. Conventional, undisputed technical context can supply missing details.
- Substantial evidence review remains a steep hill for appellants to climb.
- Minor claim variations can drive divergent outcomes in IPRs, even when the same prior art is in play.
For companies managing large authentication, identity, or platform ecosystems, the decision also highlights the continuing vulnerability of older software patents to careful prior art analysis grounded in industry practice.
As Charles Gideon Korrell notes, this case fits comfortably within the Federal Circuit’s broader effort to balance fairness to patent owners with deference to the PTAB’s technical fact-finding. The court did not expand Board authority, but it did confirm that reasoned inferences, grounded in the record and common sense, will be upheld.