Articles

CPG Still Treats Patents as Optional. AI Speed and Discretionary Denial Made That a Bad Bet

Date: September 10, 2026

Consumer packaged goods companies know how to spend on a brand. They know how to lock a formula in a vault. They are much slower to treat a patent as an operating asset.

That habit was always expensive. It is more expensive now.

Two facts changed the math. First, generative AI tools compressed the time it takes a follower to reformulate, restage a pack and put a near-SKU on the shelf. Speed of innovation is no longer a private advantage. It is a shared capability. Second, the U.S. Patent and Trademark Office spent 2025 rewriting how inter partes review (IPR) and post-grant review (PGR) get instituted. The cheap, almost-automatic Patent Trial and Appeal Board (PTAB) challenge that in-house counsel used as a reason not to file is no longer the system they remember.

Patents were already the most overlooked asset in CPG. They are now one of the few remaining ways to buy time now that a competitor can copy the idea in a quarter.

This is not a claim that every snack needs a patent, or that IPR is dead. Institution is still discretionary and fact-dependent. It is an observation that CPG general counsel who still budget patents like a pharma curiosity, and who still assume the PTAB will take every petition, are missing the opportunity to create important value for their companies.

The CPG stack that leaves patents off the page

Ask a CPG legal team what they protect and you will hear a familiar list: the trademark consumers say out loud, the trade dress on the shelf, the formula nobody is supposed to see, the copacker NDA, the celebrity license. Patents, if they appear, appear as a line item for a "platform" technology or a one-off gadget in the cap.

That is not stupidity. It is path dependence.

CPG in-house counsel skip patents for reasons that used to be rational:

They think the real property is the brand. A name and a look can last decades. A composition claim looks like a science project that a flavor house will design around in a season.

They think disclosure is the enemy. A patent has to teach the invention. A formula in the vault does not. In a category where reverse engineering is old news, teaching the art feels like paying to educate the copycat.

They think patents are for pharma and electronics. Those industries live on exclusivity windows and claim charts. CPG lives on slotting fees, promotions and velocity. The internal client is a brand manager, not a VP of IP. The budget sits in marketing.

They think the SKU will be gone before the patent issues. Eighteen months of prosecution against an 18-week seasonal set feels like the wrong clock.

They think any patent they do get will be IPRed. For a decade, the Board was the place accused infringers went to put a patent on ice. If the asset dies in a year at the PTAB, why spend the filing fees?

Each of those points still has a piece of truth. None of them is a strategy after AI made copying cheap and the USPTO made PTAB institution harder.

AI did not make CPG innovation less valuable. It made unowned innovation cheaper to steal.

A model can propose a sodium cut, a clean-label emulsifier swap, a resealable geometry or a "better for you" claim set in an afternoon. Your own R&D team can use that. So can the number-two brand, the private-label buyer and a well-funded disruptor with no plant of its own.

When everyone can search the same ingredient space faster, the advantage is not "we thought of it first" in a lab notebook. The advantage is the legal right to stop the second mover from selling it. That is what a patent is for. Trade secrets still matter for the process the box cannot teach. Trademarks still matter for the name on the box. Neither one stops a competitor who independently arrives at a close composition and a different brand.

The companies that treat AI as a reason to skip patents have the causation backwards. If the follower can close the gap in weeks, the first mover needs exclusion more, not less. A trade secret that reverse engineering plus a model can reconstruct is a shorter-lived secret. A brand without a composition or process claim is a brand competing on spend. A patent that actually maps to the SKU is a period of time in which the copycat has to design around you, license you or wait.

That is the overlooked asset. Not a stack of unexamined provisional patent applications. A human-conceived composition, process or article, claimed at a scope the specification supports, sitting on the product that makes the money.

The IPR assumption in-house still uses is out of date

The other half of the old CPG story was: even if we get a patent, a petitioner will knock it out. That story depended on a particular PTAB.

On February 28, 2025, the USPTO rescinded former Director Kathi Vidal's June 21, 2022 memorandum on discretionary denials in AIA trials with parallel district-court litigation. That 2022 memo had told the Board not to deny institution under Fintiv in several recurring situations, including when a petition presented "compelling evidence of unpatentability," when the parallel case was at the ITC or when the petitioner gave a Sotera-style stipulation. After the rescission, the Office told parties to go back to Board precedent, including Apple Inc. v. Fintiv, Inc. and Sotera Wireless, Inc. v. Masimo Corp. Portions of later decisions that had relied on the 2022 memo are not binding or persuasive.

That was the first shoe. It restored Fintiv as a live screen, not a screen with three automatic off-ramps.

On March 26, 2025, Acting Director Coke Morgan Stewart issued Interim Processes for PTAB Workload Management. Institution decisions in IPR and PGR were split. Discretionary considerations go first, to the Director in consultation with at least three PTAB judges. If discretionary denial is appropriate, the Director denies and the petition never reaches a merits panel. If it is not, the petition is then referred for a merits and statutory institution decision. Patent owners got a separate discretionary brief. Petitioners got an opposition. The memo told parties they could address the usual Board precedent (Fintiv, General Plastic, Advanced Bionics) and a list that included prior validity adjudications, changes in law, strength of the challenge, reliance on expert testimony, "settled expectations of the parties, such as the length of time the claims have been in force," and compelling economic, public-health or national-security interests. Workload and ex parte appeal pendency were also on the table.

On October 17, 2025, Director John A. Squires went further. Effective October 20, 2025, the Director determines whether to institute IPR and PGR trials. After reviewing discretionary considerations, the merits and non-discretionary considerations, with at least three PTAB judges consulted, the Director issues a summary notice granting or denying institution. Instituted trials still go to a three-member PTAB panel. Routine institution decisions are summary notices. The Director's own memo recorded that the Office had already issued more than 580 decisions under the March interim processes.

Read those three documents in order and the CPG assumption breaks. IPR is still available. 35 U.S.C. § 314(a) still says the Director may not institute unless the petition shows a reasonable likelihood of prevailing on at least one claim. The statute still uses "may," not "shall." What changed is the gate. Discretionary denial is no longer a side issue that a strong petition could brush past under the 2022 memo. It is a Director-level threshold. Parallel litigation, serial petitions, already-adjudicated claims and the age of the patent ("settled expectations") are in the first conversation, not the last.

That is not a holding that every CPG patent will survive. It is a holding that a well-built CPG patent is now much harder to use the PTAB as a cheap reset button against. In-house counsel who still price patents as if 2018 institution practice were in force are using a stale discount rate.

A few limits still remain:

Discretionary denial is not a guarantee. A young patent, a weak parallel case, a petition that is the first challenge and not a follow-on, or a record that does not support "settled expectations" can still be instituted. The March 26 memo is interim and workload-driven. The October 17 memo is Office process, not a Federal Circuit opinion. District court and ITC validity challenges did not disappear. Enablement, obviousness and eligibility still apply in court. A patent that merely recites "use a model to pick a snack formula" has other problems besides IPR.

The point is narrower, and it is the one CPG counsel actually needs: the expected value of a real patent went up because the expected cost of killing it at institution went up.

Why that value shows up in CPG first

Pharma already knew patents were the product. Software already learned, the hard way, that a claim count is not a moat. CPG is the industry that still behaves as if exclusivity were optional because the shelf used to move slowly enough for brand spend to do the work.

AI removed the slow. A private-label reformulation that used to take a development cycle can be proposed in a sprint. A pack that used to need a design studio can be iterated overnight. The companies that will take price in that market are the ones that can say: you may not sell that composition, that process or that article.

The overlooked move is not "file more provisionals." It is to put patents on the same footing as trademarks in the brand plan:

Identify the SKU that will still be on shelf in three years. Ask whether a competitor with a model and a copacker could sell a close version without using your name. If yes, ask whether a human on your team conceived a specific composition, process or pack structure that you can claim and enable. If yes, file. If the only differentiator is a slogan and a vaulted recipe that the label already half-discloses, you do not have a patent problem. You have a product problem.

Age of the patent now has a PTAB meaning. "Settled expectations" in the March 26 memo includes how long the claims have been in force. That is an argument for filing earlier than the CPG instinct (wait until the SKU is a hero), and for keeping the family alive long enough that a later petitioner is attacking an asset the market has already organized around. It is also an argument for not treating a six-year-old composition patent as dead weight in a sale.

Parallel litigation is back in the institution analysis as a first-class Fintiv issue, not as something a Sotera stipulation or "compelling merits" automatically overrode under the 2022 memo. For a CPG company that actually enforces, that can mean the patent is more useful in district court because the accused infringer cannot count on a stay-and-kill at the Board. For a CPG company that never enforces, it means nothing. The asset only reprices if you are willing to use it.

What in-house is still getting wrong

The GC who says "we are a brand company" is describing a budget, not a legal conclusion. Brand without exclusion is rented attention. The GC who says "patents require disclosure" is right about the statute and wrong about the alternative. The alternative is a secret that a model-assisted copacker can recreate from the box, plus a trademark the copycat simply does not use.

In-house counsel who says "patent prosecution is too slow for CPG" is comparing the wrong clocks. You do not need the patent to issue before the first seasonal run. You need it in force when the copycat arrives, which is now sooner thanks to AI. A provisional on the composition you are actually scaling is not theater if the later nonprovisional can support the claims.

An attorney who says "we'll get IPRed" is reciting 2018. Ask the litigation team whether they have read the USPTO’s February 28 rescission, the March 26 interim processes and the October 17 Director-institution memo. If the answer is no, the patent budget is being set with a map of a Board that no longer exists.

A company lawyer who says "our formulations are obvious" might be right about some of them. Then do not file those. File the ones with a technical difference you can enable: a stable emulsion at a specified pH, a process window that survived plant scale-up, a pack structure that solved a real barrier or opening problem. Obviousness is a reason to claim better, not a reason to own nothing.

Questions for CPG general counsel

  • Which current or pipeline SKUs would still be worth copying if the copier could not use your brand? If the list is empty, patents will not save you. If it is not empty, you are leaving exclusion on the table.

  • For each of those SKUs, is the differentiator a composition, a process, an article of manufacture or only marketing? Patents attach to the first three.

  • Who is the natural person who conceived that differentiator, and can they describe it without the model transcript?

  • Are you still discounting patents because you assume IPR institution is the default? Have you updated that assumption against the 2025 Office process?

  • If you were sued on a competitor's CPG patent tomorrow, would you still treat an IPR petition as a reliable first move, or would discretionary denial be the first memo you asked for?

  • In a sale of the brand, can you map a patent claim to revenue, or only a trademark and a recipe folder?

Questions for buyers of CPG companies

  • Do not pay extra for "AI-enabled innovation" that is not claimed. Pay extra for claims that map to SKUs and that a petitioner now has to get past a Director-level discretionary screen.

  • Ask when the key patents issued. Age is no longer only a remaining-term question. It is a settled-expectations question at institution.

  • Ask whether the seller has ever enforced. A patent family that has never been asserted is still property. It is weaker evidence that the company knows how to use the new PTAB posture.

  • Read the claims against the product, not against the deck.

About the author

Jeffrey R. Schell is a registered patent attorney and Managing Partner of the Mountain West practice at Whiteford, where he counsels technology companies on intellectual property strategy, AI governance and venture growth. A former multi-exit software founder and trained machine learning engineer, he advises companies from startup through exit on building defensible IP positions. That combination of engineering and transactional experience maps directly onto the questions raised here: whether a CPG company's real differentiator is a claimable composition or process, how the USPTO's 2025 overhaul of PTAB discretionary denial changes the economics of filing, and why AI-accelerated copying makes patents — not just brand spend — the asset worth defending.

aBOUT wHITEFORD

Whiteford provides comprehensive business law and litigation services to clients ranging from innovative start-ups to middle market companies to Fortune 100 enterprises. With a growing footprint of East Coast offices from New York to Florida, and a new office in Denver, Colorado, we serve clients regionally, nationally and internationally.


Source notes

  1. USPTO, USPTO rescinds memorandum addressing discretionary denial procedures (Feb. 28, 2025) https://www.uspto.gov/about-us/news-updates/uspto-rescinds-memorandum-addressing-discretionary-denial-procedures
  2. USPTO, Interim Procedure for Discretionary Denials in AIA Post-Grant Proceedings with Parallel District Court Litigation (June 21, 2022) (rescinded Feb. 28, 2025).
  3. Apple Inc. v. Fintiv, Inc., IPR2020-00019, Paper 11 (PTAB Mar. 20, 2020) (precedential).
  4. Sotera Wireless, Inc. v. Masimo Corp., IPR2020-01019, Paper 12 (PTAB Dec. 1, 2020) (precedential as to § II.A).
  5. Coke Morgan Stewart, Interim Processes for PTAB Workload Management (Mar. 26, 2025) https://www.uspto.gov/sites/default/files/documents/InterimProcesses-PTABWorkloadMgmt-20250326.pdf
  6. USPTO, USPTO issues new interim process concerning institution of AIA proceedings (Mar. 26, 2025) https://www.uspto.gov/subscription-center/2025/uspto-issues-new-interim-process-concerning-institution-aia-proceedings
  7. General Plastic Indus. Co. v. Canon Kabushiki Kaisha, IPR2016-01357, Paper 19 (PTAB Sept. 6, 2017) (precedential as to § II.B.4.i).
  8. Advanced Bionics, LLC v. MED-EL Elektromedizinische Geräte GmbH, IPR2019-01469, Paper 6 (PTAB Feb. 13, 2020) (precedential).
  9. John A. Squires, Director Institution of AIA Trial Proceedings (Oct. 17, 2025) https://www.uspto.gov/sites/default/files/documents/Director_Institution_of_AIA_Trial_Proceedings.pdf
  10. USPTO, Director institution of AIA trial proceedings (Oct. 17, 2025) https://www.uspto.gov/subscription-center/2025/director-institution-aia-trial-proceedings
  11. 35 U.S.C. § 314(a)–(c) https://www.law.cornell.edu/uscode/text/35/314
  12. 35 U.S.C. § 324(a) https://www.law.cornell.edu/uscode/text/35/324
The information contained here is not intended to provide legal advice or opinion and should not be acted upon without consulting an attorney. Counsel should not be selected based on advertising materials, and we recommend that you conduct further investigation when seeking legal representation.