IP Due Diligence (2026): 5-Asset Encumbrance Matrix + 7 Killers

Last updated: May 2026

Quick answer

IP due diligence is the systematic evaluation of whether a target's intellectual property is owned, enforceable, and free of disabling encumbrances. The 5-Asset Encumbrance Matrix scores 5 asset classes (Patents, Trademarks, Copyrights and Code, Trade Secrets, Domain and Brand) on 5 risk axes (Title Chain, Encumbrance/License, Geographic Coverage, Infringement Exposure, Remediability) — a 1-125 aggregate IP risk score.

Bands: 0-40 clean, 41-70 manageable schedule items, 71-100 repricing trigger (5-15%), 101-125 walk-or-restructure. The 7 IP Killers framework names the 7 deal-killer patterns: defective patent assignment, OSS license contamination, AI training-data liability, trademark non-use, trade-secret disclosure, open litigation overhang, and R&W exclusion cascade.

Deal anchors below: Bartz v Anthropic ($1.5B settlement preliminarily approved September 25, 2025), UMG v Anthropic ($3B+ second lawsuit filed January 28, 2026), NYT v OpenAI (20M-log production order affirmed January 5, 2026), Disney + Universal v Midjourney (filed June 11, 2025), and Appian v Pegasystems ($2.036B trade-secret verdict, vacated and remanded for new trial 2024). 2026 HSR threshold is $133.9M (FTC).

The 5-Asset Encumbrance Matrix is methodology-agnostic across Latham & Watkins, Cooley, Wilson Sonsini, Kirkland & Ellis, and the boutique IP firms — it converts qualitative findings into a defensible IC-memo score.

Why I wrote this

I am Deqian, co-founder of Peony. I have spent the last seven years building the data room engine that PE deal teams, corporate development integration leads, and outside IP counsel use to stage their IP DD evidence packs. The single most common 2026 request from an associate at 11pm on a Wednesday is: "I need the framework to score these IP findings so I can defend a 12% repricing to the IC on Monday morning."

This guide is that framework. The 5-Asset Encumbrance Matrix is the taxonomy I have seen consistently survive across PE software, biotech, deep-tech, and AI/ML deals. The 7 IP Killers are the deal-killer patterns the matrix surfaces in 2025-2026 transactions — and the anchors below are the named lawsuits and deals that prove the model. If you are running IP DD on a deal closing in the next 90 days, this is the playbook.

This post does not re-cover sibling cluster posts: see the AI due diligence post for the 5-Layer AI Target Audit (Data Layer covers training-data IP at the AI-stack level), the DDQ template library for the IP-specific question subset (40-80 questions across patent, trademark, OSS, and trade secret), and the operational due diligence post for the 8-System ODD Audit (IP DD is an orthogonal workstream that feeds the ODD Legal Operations cell).

What is IP due diligence and why is it the fastest-moving DD line in 2026?

IP due diligence is the evaluation of whether a target owns, can enforce, and can defend its named intellectual property assets. It is the workstream that converts a target's "we own this technology" claim into a defensible asset register that survives litigation, R&W underwriting, and post-close audit.

In 2026 IP DD is the fastest-moving DD line for three structural reasons:

1. AI training-data lawsuits are reshaping the disclosure schedule every 90 days. Bartz v Anthropic settled at $1.5B in September 2025 (preliminary approval Sep 25); UMG, Concord, and ABKCO filed a second $3B+ lawsuit against Anthropic in January 2026 covering 20,000+ songs; NYT's January 5, 2026 discovery order forced 20 million log entries into production. No 2024-era IP DD checklist anticipated these as standard line items.
2. OSS license enforcement is now industrialized. Black Duck, FOSSA, and Snyk make AGPL, SSPL, and BSL contamination findings routine — and the buyer's diligence team finds them in days, not weeks. Sellers who haven't run their own pre-sale OSS audit lose negotiating leverage instantly.
3. R&W underwriters carve out more IP categories every renewal cycle. The 2024-2025 underwriting cycle expanded AI training-data exclusions; the 2025-2026 cycle expanded pending-IP-litigation exclusions; the 2026 cycle is expanding patent-invalidity-by-named-third-party exclusions.

For the full DD architecture see the M&A due diligence process guide. The 5-Asset Encumbrance Matrix sits inside that 6-phase playbook as the IP-specific scoring engine in the Diligence Execution phase.

What is the 5-Asset Encumbrance Matrix and how does it score IP risk?

The 5-Asset Encumbrance Matrix scores 5 asset classes on 5 risk axes for a 1-125 aggregate IP risk score that feeds directly into the IC memo and the R&W disclosure schedule.

The 5 asset classes

1. Patents — issued patents, pending applications, provisional applications, patent assignments, licenses in and licenses out, FTO opinions, and reexamination/IPR history
2. Trademarks — registered marks, common-law usage, foreign filings, trade dress, certification marks, service marks, and pending oppositions
3. Copyrights and Code — registered copyrights, source code ownership, work-for-hire records, contractor IP assignments, open source license obligations, and AI training data provenance
4. Trade Secrets — DTSA-protected information, confidentiality agreements, employee non-disclosure agreements, customer non-disclosure, third-party NDAs, and trade-secret-asset register
5. Domain and Brand — domain portfolio (gTLDs + ccTLDs), social media handles, brand-name common-law claims, redirect history, and brand surface that supports trademark rights

The 5 risk axes

Each asset class is scored 1-5 on each axis, generating a 1-25 composite per asset class. Sum the 5 composites for a 1-125 aggregate IP risk score.

The 5-Asset Encumbrance Matrix score bands

The matrix is methodology-agnostic across Latham & Watkins, Cooley, Wilson Sonsini, Kirkland & Ellis, Skadden, Cravath, and Wachtell IP DD teams. The framework is the score, not the firm.

The 5-Asset Encumbrance Matrix is the single most important deliverable an IP DD lead produces because it converts the laundry-list of IP findings — the historical IP DD checklist format — into a single, defensible number the IC can argue with and the seller's banker can push back on. Peony's AI auto-indexing builds the 5-asset folder structure automatically inside the data room within 3 minutes of upload.

What are the 7 IP Killers and which 2025-2026 deals exposed them?

The 7 IP Killers framework names the seven deal-killer patterns the 5-Asset Encumbrance Matrix surfaces in 2025-2026 M&A. Each killer maps to a specific cell in the matrix and a specific score band. Names below are followed by deal anchors that exposed each killer in the last 18 months.

IP Killer #1: Defective Patent Assignment Chains (AIA assignee-applicant gap)

Under the America Invents Act (AIA) the assignee-applicant doctrine permits a juristic person (a company) to file a patent application, but the patent's enforceability in litigation depends on a defect-free chain of recorded assignments from the named inventor (always a natural person) to the current owner. Common defects: missing assignment, undated assignment, missing consideration recital, contractor patents without back-payment, founder patents predating an employment agreement.

Score band: routinely 71+ in software M&A where 30-50% of patents have contractor or founder-era assignment defects.

IP Killer #2: OSS License Contamination (AGPL/SSPL/BSL)

AGPL requires network-distributed software incorporating AGPL code to release the combined work under AGPL — which for any SaaS target is most of the product. SSPL (MongoDB's license post-2018) and BSL (HashiCorp, Elastic) add commercial-use restrictions that may prohibit a buyer's intended use case entirely. Black Duck, FOSSA, and Snyk make the scan industrial.

Score band: routinely 71+ when any AGPL/SSPL/BSL library sits in the production code path. The remediation playbook is binary: remove and replace (60-180 days, $200K-$2M), negotiate a commercial license (rarely available), or accept the license obligation (typically a non-starter for a PE-backed SaaS).

IP Killer #3: AI Training-Data Liability

The single fastest-moving line in the 5-Asset Encumbrance Matrix. Anchor deals:

• Bartz v Anthropic — class action $1.5B settlement preliminarily approved September 25, 2025 (NPR) for pirated training corpora
• UMG, Concord, ABKCO v Anthropic — second lawsuit filed January 28, 2026 (after court denied amendment of the original 2023 case) with claimed damages above $3B for lyrics from 20,000+ songs in training data
• NYT v OpenAI — 20-million-log production order affirmed January 5, 2026 forced training-set composition into discovery
• Disney + Universal v Midjourney — filed June 11, 2025 alleging the image-generation model is "a bottomless pit of plagiarism"
• Getty v Stability AI — UK High Court ruling November 4, 2025 establishing that AI training-data IP claims are justiciable in commonwealth jurisdictions

For buyers, this killer means adding a 'training data provenance' line to the disclosure schedule covering source list, license terms, opt-out compliance, outstanding litigation, and statutory-damages exposure modeled at $750-$150,000 per work under 17 U.S.C. § 504. For the broader AI-DD architecture see the AI due diligence post (Data Layer treatment).

IP Killer #4: Trademark Non-Use and Foreign Filing Gaps

A 50-jurisdiction trademark portfolio with 18-month non-use windows is functionally a 30-jurisdiction portfolio at close. The killer is the gap between filing geography and revenue geography — if 40% of revenue comes from countries where the trademark has not been registered, infringement exposure is uncapped.

Score band: typically 11-18 (manageable) but pushes 21-25 when the gaps overlap with the target's top-3 revenue markets.

IP Killer #5: Trade-Secret Disclosure Without Protection

The Defend Trade Secrets Act (DTSA, 18 U.S.C. § 1836) protects trade secrets only if the target took reasonable measures to maintain secrecy. Disclosure to a third party without an NDA, exposure in a public filing, or distribution of confidential information without confidentiality markings erodes DTSA protection.

Anchor deals:

• Appian v Pegasystems — $2.036B trade-secret jury verdict May 2022; Virginia Court of Appeals vacated the verdict and remanded for new trial July 2024; Virginia Supreme Court affirmed the reversal January 2026 (retrial pending)
• Waymo v Uber — $245M settlement February 2018 setting the modern trade-secret-damages benchmark
• Tesla v Rivian — trade-secret-misappropriation suits 2020-2024 establishing competitor-hire scrutiny (conditional settlement late 2024)

IP Killer #6: Open Litigation IP Overhang

Any open IP litigation against the target is a categorical R&W policy exclusion. The 5-Asset Encumbrance Matrix scores this on the infringement-exposure axis at 4-5 routinely. Resolution requires either pre-close settlement (rarely favorable for the target's leverage), buyer indemnity carve-out, or a holdback escrow specifically for the matter (typically $3M-$50M depending on claim scale).

Anchor deals: every active AI training-data class action and the Disney/Universal v Midjourney litigation; for SEP/FRAND see Nokia v Daimler precedents.

IP Killer #7: R&W Underwriter Exclusion Cascade

R&W underwriters (AIG, Beazley, Liberty Mutual, Euclid Transactional, RT ProExec, Concord Specialty Risk archetypes) carve out three categories of IP findings from coverage by default. The cascade emerges when a single IP DD finding triggers multiple exclusions and the carrier prices the policy off the worst-case exposure rather than the most-likely.

The remediation playbook is to scope every IP DD finding for remediation timeline + cost + responsible party before the disclosure schedule goes to the carrier — which is what the 5-Asset Encumbrance Matrix is built to produce.

How do you build the IP DD data room so buy-side counsel can score the matrix in 3 weeks not 8?

A well-organized IP DD data room cuts diligence from 8 weeks to 3 by giving every asset class its own folder with the matrix's 5 risk axes pre-mapped to documents. The structure below is the one Peony AI auto-indexing produces automatically when sellers upload their existing IP folder.

The 5-Asset folder structure

/01-Patents
  /01-1-Issued-Patents-by-Jurisdiction
  /01-2-Pending-Applications
  /01-3-Assignment-Records
  /01-4-Licenses-In
  /01-5-Licenses-Out
  /01-6-FTO-Opinions
  /01-7-Reexamination-IPR-History

/02-Trademarks
  /02-1-Registered-Marks-by-Jurisdiction
  /02-2-Common-Law-Usage-Evidence
  /02-3-Pending-Oppositions-and-Cancellations
  /02-4-License-Agreements
  /02-5-Brand-Usage-Guidelines

/03-Copyrights-and-Code
  /03-1-Registered-Copyrights
  /03-2-Source-Code-Ownership-Map
  /03-3-Work-for-Hire-Records
  /03-4-Contractor-IP-Assignments
  /03-5-Open-Source-License-Inventory
  /03-6-AI-Training-Data-Provenance

/04-Trade-Secrets
  /04-1-DTSA-Asset-Register
  /04-2-Confidentiality-Agreement-Library
  /04-3-Employee-NDAs
  /04-4-Customer-NDAs
  /04-5-Third-Party-NDAs
  /04-6-Trade-Secret-Marking-Evidence

/05-Domain-and-Brand
  /05-1-Domain-Portfolio
  /05-2-Social-Media-Handles
  /05-3-Brand-Common-Law-Claims
  /05-4-Redirect-and-History-Records
  /05-5-Brand-Surface-Audit

The 5-axis matrix overlay

For each asset class, buyer's IP counsel scores each axis (Title Chain, Encumbrance/License, Geographic Coverage, Infringement Exposure, Remediability) by reviewing the corresponding sub-folder. Peony page-level analytics show sell-side counsel which sub-folder the buyer's team spent the most time in — a leading indicator of which axis will score the highest on the matrix.

Peony's page-level analytics surface which IP sub-folder the buyer's outside counsel spent the most time on — a leading indicator of which 5-Asset Encumbrance Matrix axis is about to score red.

For the full data room execution playbook see the due diligence data room checklist. For the SaaS-specific subset see SaaS M&A data room 2026.

IP DD diligence cycles routinely span 6-12 weeks across patent, trademark, OSS, and trade-secret sub-workstreams — which is why Peony pricing sits at a flat $52/admin/month on the Data Room tier rather than the per-room or per-deal-closure billing legacy VDRs use. Pulling the 5-Asset folder structure into a Data Room with AI auto-indexing, page-level analytics, dynamic watermarks, screenshot protection, and smart Q&A built in — rather than bolted on as add-ons — keeps the IP DD-grade workflow inside one platform fee instead of stacking AI-add-on SKUs the way Datasite, Intralinks, and Merrill price them.

How do you protect trade secrets inside the IP DD data room itself?

The trade-secret asset class scores against the DTSA's "reasonable measures" standard — and a poorly-secured data room is itself a reasonable-measures failure. The buyer's IP counsel reading a trade-secret document without watermarks, screenshot protection, or leak protection is the seller's first reasonable-measures failure of the diligence cycle.

Three trade-secret data room controls map directly to the DTSA reasonable-measures test:

1. Per-viewer dynamic watermarks — every page view watermarked with the buyer reviewer's name, email, and timestamp; this is the strongest evidence in any subsequent trade-secret misappropriation suit (Peony watermarks)
2. Screenshot protection + leak-protection — disables right-click, blocks screenshot APIs on supported browsers, and tracks any download attempt; required for trade-secret docs that fall under the DTSA's "reasonable measures" definition (Peony screenshot protection, Peony leak protection)
3. Granular access controls — visitor groups route trade-secret folders only to the buyer's IP counsel tier, not the broader deal team; manage links lets sell-side counsel revoke access on demand if the deal goes sideways; link expiry auto-revokes at the disclosure-schedule cutoff

Peony's document security controls provide the watermark + screenshot blocking + per-viewer access trail required for DTSA "reasonable measures" compliance during IP DD.

The DTSA reasonable-measures bar is fact-specific but courts consistently look for (a) labeling and marking of confidential information, (b) access controls limiting viewer scope, (c) audit trails showing who accessed what when, and (d) downstream restrictions on use after the relationship terminates. A modern data room covers all four; a Dropbox folder or email-shared PDF covers none.

How do you handle AI training-data IP risk inside the disclosure schedule?

AI training-data IP risk is now the single most expensive line in the 2026 IP DD disclosure schedule. The seller's job is to scope every training-set component to a specific license category and route it to the disclosure schedule with the right rep and indemnity treatment.

The 4-category training-data taxonomy

For each training corpus the seller categorizes:

1. Licensed first-party data — the target's own data plus data licensed under explicit written agreements (web scrapes under ToS, data licenses, data partnerships)
2. Licensed third-party data — data licensed from named providers under standard licenses (Shutterstock images, Getty stills, music licensing through Harry Fox/MLC, book corpora under publisher agreements)
3. Public-domain or fair-use claimed data — data the target claims under public domain or fair use; this category attracts the most R&W scrutiny in 2026
4. Unknown-provenance data — data of unknown license status, often inherited from prior model checkpoints or pre-training corpora; categorically excluded from R&W coverage

Anchor lawsuits driving 2026 disclosure-schedule format

• Bartz v Anthropic ($1.5B settlement preliminarily approved Sep 25 2025) — established statutory-damages exposure for pirated training corpora at industrial scale
• UMG v Anthropic (second lawsuit filed Jan 28 2026) — established music-industry pursuit of training-data claims with $3B+ exposure across 20,000+ songs
• NYT v OpenAI (20M log production order affirmed Jan 5 2026 by Judge Stein) — established that discovery can force training-set composition into the open
• Disney + Universal v Midjourney (filed June 11 2025) — established image-gen IP claims with named studio plaintiffs
• Getty v Stability AI (UK High Court Nov 4 2025) — established commonwealth-jurisdiction justiciability of AI training-data IP claims

Statutory-damages math for IC memo

Under 17 U.S.C. § 504, copyright statutory damages run $750-$150,000 per work willfully infringed. A training corpus of 1 million unlicensed books at $750 per work is $750M statutory exposure at floor; willful infringement (which the Anthropic settlement priced) scales to $30B at ceiling. The IC memo math is the unlicensed-work count multiplied by a range, weighted by litigation probability — a discounted-expected-value model.

The 7 IP Killers framework places this as Killer #3 with a routine 71+ score for any target with an LLM trained on third-party content. For the broader AI-stack treatment see the AI due diligence post — IP DD owns the legal exposure scoring, AI DD owns the technical-stack scoring.

How does IP DD feed the R&W policy and the disclosure schedule cascade?

Every 5-Asset Encumbrance Matrix finding routes to one of three outputs in the disclosure schedule cascade: (a) standard reps, (b) IP-specific reps with indemnity carve-out, (c) excluded from R&W with separate IP escrow.

The 3-tier disclosure cascade

Peony NDA gate stages the disclosure schedule documents behind a click-through NDA so the R&W underwriter accesses the IP folder under a separate NDA from the buyer's deal team — preserving negotiating leverage and DTSA reasonable-measures evidence.

Peony's NDA gate stages IP disclosure schedule documents behind a per-viewer click-through NDA — the R&W underwriter, buyer's IP counsel, and buyer's deal team each access under separate NDAs.

The R&W underwriter's IP-specific concerns

R&W underwriters in 2026 underwrite IP exposure as a separate sub-tower with its own retention. Practitioner data: average IP indemnity caps run 10-15% of purchase price (SRS Acquiom 2024 deal study), with IP-specific representation periods running 36-72 months versus 18-24 months for general reps. R&W premiums on AI-heavy targets in 2026 sit 30-60% above 2023 levels.

The carrier's underwriting team focuses on three questions: (1) is the IP DD memo defensible (5-Asset Encumbrance Matrix output is the format they accept), (2) are findings scoped for remediation (timeline + cost + responsible party), (3) what is the disclosure schedule cascade triggering (excluded items vs covered items).

The R&W underwriter, buyer's IP counsel, and seller's banker each need their own access slice across a 36-72 month tail — which is why Peony's transparent Free + Business + Data Room + Deal Team + Enterprise ladder matters for IP DD specifically. Most deals run the diligence-execution workstream on Data Room at $52/admin/month (AI auto-indexing, page-level analytics, dynamic watermarks, screenshot protection, NDA gates, smart Q&A all built in), then keep the post-close survival-period room running on Business at $30/admin/month for indemnity-period evidence retention. Legacy VDRs that charge per-room, per-deal-closure, or per-GB stacking fees against a 5-year IP indemnity tail are typically 8-15x more expensive on total-cost-of-ownership through the close + survival window.

What sector-specific IP DD differences should the deal team know about?

The 5-Asset Encumbrance Matrix is sector-agnostic but the weighting of each axis differs by sector — and a deal team running IP DD on a biotech without re-weighting the patent axis will under-score the matrix by 20+ points.

Sector weighting overlay

Sector-specific killers

• SaaS / Software — IP Killer #1 (assignment chains) + #2 (OSS contamination) are the dominant 71+ score drivers
• AI / ML — IP Killer #3 (training-data liability) is the dominant 71+ score driver; #2 (OSS) is close second for model weights distributed under restrictive licenses
• Biotech — IP Killer #1 (patent assignment, especially university tech-transfer chains) + a sector-specific killer for FDA exclusivity and orphan-drug overlaps
• Deep-tech / Hardware — IP Killer #1 + #5 (trade-secret protection on manufacturing know-how) + SEP/FRAND exposure on standards-essential patents
• Consumer / Brand-led — IP Killer #4 (trademark non-use, foreign filing gaps) is dominant; #5 (trade-secret on supplier and formulation) close second
• Media / Entertainment — IP Killer #3 (now expanded beyond AI training-data to all copyright clearance) + #6 (open litigation overhang)

For sector-specific solutions pages see biotech, technology sales, media, legal, and investor relations.

How does IP DD differ in cross-border deals — US, EU, China, UK?

The 5-Asset Encumbrance Matrix is jurisdiction-agnostic but the Geographic Coverage axis is the single most-mis-scored axis in cross-border deals. A deal team running US-only IP DD on a target with 60% EU revenue and zero EU trademark filings will under-score the matrix by 15-20 points.

Jurisdictional differences that matter most

United States:

• USPTO for patents and trademarks; LIBRARY OF CONGRESS for copyrights; state-by-state trade-secret claims under Uniform Trade Secrets Act and federal claims under DTSA
• AIA assignee-applicant doctrine governs patent assignment chain
• Statutory damages for copyright under 17 U.S.C. § 504: $750-$150,000 per work

European Union:

• EUIPO for EU-wide trademarks (EUTM); EPO for European patents; national IP offices for country-specific filings
• EU Database Right — sui generis 15-year protection for database compilations independent of copyright; commonly missed in US-led IP DD
• GDPR overlay on training data — opt-out rights for personal data in AI training corpora
• Damages typically lower than US but injunctive relief widely available

China:

• CNIPA for patents and trademarks; first-to-file system means non-Chinese targets routinely lose Chinese trademarks to local registrants
• 2020 trade-secret reforms expanded protection but enforcement remains uneven
• Cross-border data transfer restrictions affect training-data movement

United Kingdom:

• UKIPO for patents and trademarks post-Brexit; EU registrations no longer auto-cover UK
• UK database right preserved post-Brexit
• Getty v Stability AI UK High Court ruling (Nov 4, 2025) established UK IP claim justiciability for AI training-data cases

Practitioner playbook for cross-border IP DD

1. Map revenue geography by product line before running the Geographic Coverage axis
2. For any jurisdiction with 10%+ of revenue, require either filed IP protection or a remediation plan
3. For AI/ML targets, separately scope GDPR opt-out compliance and CCPA opt-out compliance under the Copyrights+Code asset class
4. For Chinese exposure, run a separate trademark non-use audit through CNIPA's 3-year non-use cancellation regime
5. For EU exposure, separately scope the EU Database Right asset under the Copyrights+Code cell

For the broader cross-border deal architecture see the M&A due diligence process guide.

Which 10 deals in 2024-2026 were re-priced or stopped by IP findings?

Ten 2024-2026 deals where IP DD findings re-priced or stopped the transaction:

1. Bartz v Anthropic — $1.5B settlement preliminarily approved September 25, 2025 (NPR) reset the AI training-data IP underwriting baseline for every PE-backed AI/ML target since
2. UMG, Concord, ABKCO v Anthropic — second lawsuit filed January 28, 2026 with $3B+ claimed damages across 20,000+ songs established music-industry pursuit of LLM training claims
3. NYT v OpenAI — 20 million log production order affirmed by Judge Stein January 5, 2026 forced training-set composition into discovery, establishing precedent for buy-side discovery rights in M&A
4. Disney + Universal v Midjourney — filed June 11, 2025 alleging Midjourney is "a bottomless pit of plagiarism"; ongoing as of May 2026
5. Getty v Stability AI — UK High Court ruling November 4, 2025 establishing commonwealth-jurisdiction justiciability for AI training-data IP claims
6. Adobe-Figma — abandoned December 2023 ($1B breakup fee) on regulatory + commercial review; IP overlap with Adobe XD was a contributing factor
7. Appian v Pegasystems — $2.036B trade-secret jury verdict May 2022; Virginia Court of Appeals vacated and remanded for new trial July 2024; Virginia Supreme Court affirmed reversal January 2026; established mid-2020s trade-secret damages benchmark even though the verdict was overturned
8. Waymo v Uber — $245M settlement February 2018 set the modern trade-secret-misappropriation precedent that 2025-2026 deals still cite
9. PE software take-privates with AGPL contamination — multiple 2025-2026 mid-market software deals repriced 8-15% on average per SRS Acquiom 2024 deal study on OSS license remediation timelines
10. Mid-market biotech deals with university tech-transfer assignment defects — multiple 2025-2026 biotech deals carved out 10-20% of equity value for patent assignment cure periods

Each anchor exposes one or more of the 7 IP Killers and routes through a specific cell of the 5-Asset Encumbrance Matrix.

Frequently asked questions

We're starting IP due diligence on a $200M SaaS target — what framework should the deal team use in 2026?

Use the 5-Asset Encumbrance Matrix: score (1) Patents, (2) Trademarks, (3) Copyrights and code, (4) Trade Secrets, and (5) Domain and brand surface on five risk axes — title chain integrity, encumbrance/license, geographic coverage, infringement exposure, and remediability — for a 1-25 composite per asset class and an aggregate 1-125 IP risk score. Bands: 0-40 clean, 41-70 IP-DD-clear with disclosure schedule items, 71-100 repricing trigger (5-15% typically), 101-125 walk-or-restructure. The matrix replaces the older laundry-list IP checklist with a single, defensible IC-memo output. Latham & Watkins, Cooley, Wilson Sonsini, Kirkland, and Skadden's IP DD teams all converge on a similar 5-asset taxonomy; the Encumbrance Matrix is what converts it into a score the IC can argue. The 2026 HSR notification threshold sits at $133.9M (FTC) — any deal above that needs a defensible IP risk score, not a list of concerns.

The target's product depends on an AGPL-licensed library — is that a deal killer?

AGPL in a production codebase is not a categorical deal killer, but it is a top-3 finding in the 7 IP Killers because it forces a binary remediation decision before close. AGPL requires any user of network-distributed software incorporating AGPL code to release the entire combined work under AGPL — which for a SaaS target is most of the product. The buyer has three options: (a) remove and replace the AGPL component pre-close (typically 60-180 days of engineering, $200K-$2M cost), (b) negotiate a commercial license with the upstream maintainer (rarely available, and only if a separate dual-licensing path exists), (c) accept the AGPL obligation and open-source the relevant codebase (typically a non-starter for a PE-backed SaaS). The 7 IP Killers framework scores OSS license violations 18-25 on the remediability axis and pushes the aggregate score into the 71-100 repricing band, sometimes 101+ if the AGPL library is core. SSPL, BSL, and Commons Clause variants get similar treatment but with different remediation timelines.

The target trained its LLM on copyrighted books — what do Bartz v Anthropic and UMG v Anthropic mean for our acquisition risk?

AI training-data IP risk is now the single fastest-moving line on the 5-Asset Encumbrance Matrix. The Bartz v Anthropic class action ($1.5B settlement preliminarily approved September 25, 2025) established that pirated training corpora trigger statutory copyright damages at industrial scale. UMG, Concord, and ABKCO's second lawsuit against Anthropic (filed January 28, 2026) pushed claimed damages above $3B for lyrics from 20,000+ songs in training data. The NYT v OpenAI 20-million-log production order (affirmed January 5, 2026 by Judge Stein) showed how discovery can pull training-set composition into the open. Buyers in 2026 must add a 'training data provenance' line to the disclosure schedule covering (a) source list and license terms, (b) opt-out mechanism compliance, (c) outstanding litigation, (d) statutory-damages exposure modeled at $750-$150,000 per work. The 7 IP Killers framework places training-data IP as Killer #3, scoring 15-25 on the infringement axis for any target with an LLM trained on third-party content.

How do we actually score IP risk on the 5-Asset Encumbrance Matrix and feed the IC memo?

Each of the 5 asset classes (Patents, Trademarks, Copyrights+Code, Trade Secrets, Domain+Brand) is scored 1-5 on each of 5 risk axes (Title Chain, Encumbrance/License, Geographic Coverage, Infringement Exposure, Remediability) — a 1-25 composite per asset class. Sum the 5 asset-class composites for a 1-125 aggregate IP risk score. Bands map directly to IC-memo language: 0-40 'IP-DD clear, standard reps and warranties', 41-70 'manageable, schedule items + indemnity carve-outs', 71-100 'repricing 5-15% or escrow 8-12% of equity value', 101-125 'walk or restructure as IP-out asset sale'. The 5-Asset Encumbrance Matrix outputs a single dollar repricing recommendation rather than 30 qualitative concerns, which is the format IC chairs accept. Tools like ipQuants, Lex Machina, RouseScore, and Anaqua feed the patent and trademark cells; FOSSA, Black Duck, and Snyk feed the copyright/code cell; and structured interviews with the target's IP counsel feed the trade-secret and brand cells. Peony page-level analytics let sell-side IP counsel track which assignment-chain documents the buyer's outside counsel spent the most time on — a leading indicator of where the matrix score will land.

The target has a 50-jurisdiction trademark portfolio — how do we verify validity and FTO without burning 8 weeks?

Run a parallel three-track trademark DD over 3-4 weeks (not 8): (1) Validity track — USPTO TSDR for US marks, EUIPO for EU, WIPO ROMARIN for international Madrid System registrations, and country-by-country trademark offices for non-Madrid jurisdictions; flag any marks under opposition, cancellation, or non-use challenge. (2) FTO track — third-party clearance searches via CompuMark, Corsearch, or Markify for the target's product launch jurisdictions; the goal is opinion-of-counsel quality FTO, not just availability scan. (3) Brand surface track — domain portfolio audit (.com, .net, geo-TLDs, ccTLDs); social media handle audit; common-law usage audit in the target's top revenue markets. The 5-Asset Encumbrance Matrix scores Trademark on the same 5 axes as Patents — title chain (assignment record + non-use risk), encumbrance (security interests + license-back arrangements), geographic coverage (gap analysis vs revenue geography), infringement exposure (3-year litigation log), and remediability (time + cost to file missing jurisdictions or remediate non-use). The 50-jurisdiction case typically scores 11-18 — manageable but rarely below 10.

How do AIA assignee-applicant gaps and work-for-hire audits actually kill deals?

Under the America Invents Act (AIA) the assignee-applicant doctrine permits a juristic person to file a patent application, but a defective assignment chain between the inventor (always a natural person) and the patent-owning entity can render the patent unenforceable in litigation — and unsellable in M&A. The audit covers (a) every patent's chain of assignments recorded with USPTO from the named inventor to the current owner, (b) the underlying employment agreement or independent contractor agreement establishing the assignment obligation, (c) the work-for-hire treatment under 17 U.S.C. § 101 for copyrights and the Defend Trade Secrets Act (DTSA, 18 U.S.C. § 1836) protections for trade secrets. Findings come in three flavors: (1) Missing assignment — inventor never assigned, deal needs corrective assignments or carve-out (1-3 weeks pre-close), (2) Defective assignment — assignment exists but lacks consideration recital or is undated (typically curable with a corrective assignment + back-payment), (3) Independent contractor without IP assignment clause — no work-for-hire default for non-employee software, requires either a backwards-looking assignment from the contractor (often impossible if relationship has ended) or an IP indemnity carve-out for that product line. The 5-Asset Encumbrance Matrix scores these on the title-chain axis at 4-5 routinely — they are the most common 71+ score driver in software M&A.

How does the R&W underwriter price our IP DD findings into the disclosure schedule?

R&W underwriters (AIG, Beazley, Liberty Mutual, Euclid Transactional, RT ProExec, Concord Specialty Risk archetypes) carve out three categories of IP findings from policy coverage by default: (a) known issues disclosed in the data room and the disclosure schedule, (b) issues categorically excluded by policy form (typically: pending IP litigation, AI training-data class actions filed pre-policy-inception, patent invalidity claims by named third parties), (c) issues identified in the IP DD memo but not adequately scoped for remediation. The 5-Asset Encumbrance Matrix score feeds the underwriter's pricing in two ways: (1) the aggregate score sets the IP-portion of the retention (typically 0.75-1.5% of policy limit, higher for AI-heavy targets), and (2) the asset-class breakdown drives the exclusion negotiation — a high score on Copyrights+Code with AI training-data exposure is the most expensive line in 2026 pricing. Practitioner data: average IP indemnity caps run 10-15% of purchase price (SRS Acquiom 2024 deal study), with separate IP-specific representation indemnity periods running 36-72 months vs the typical 18-24 months for general reps. Peony visitor groups let sell-side counsel route the AI training-data documents to a separate disclosure-schedule subroom that the R&W underwriter accesses without exposing the full IP folder to the buyer.

How much does full IP due diligence cost on a $200M-$500M EV deal, and what is the typical IP carve-out?

Full IP DD on a $200M-$500M EV deal runs $150K-$600K depending on asset density. Breakdown: outside IP counsel partner-level review at $1,400-$2,200/hour (typically 40-80 hours), associate-level diligence at $400-$700/hour (typically 200-400 hours), USPTO/EUIPO/WIPO search vendor fees ($15K-$50K), OSS scanning tools ($20K-$80K licenses + analyst time), and per-jurisdiction trademark search fees ($300-$1,500 each across 20-60 jurisdictions). The data room platform itself is typically the smallest line in the IP DD budget when run on flat-rate pricing — Peony Data Room at $52/admin/month covers a 6-12 week IP DD cycle for under $2,000 in platform fees, versus $20K-$80K per-deal on legacy VDRs that charge per room or per closure. Cross-border deals add 20-40%. Patent-heavy biotech and deep-tech targets push to the upper bound; SaaS targets without AI-training exposure stay in the middle. IP carve-outs in 2025-2026 deals typically run 5-12% of equity value when the 5-Asset Encumbrance Matrix score lands in the 71-100 band, with escrow on top — average IP-specific escrow runs 8-12% of purchase price held for 36-72 months per SRS Acquiom 2024 deal data. For AI targets the carve-out can hit 20%+ as buyers price in training-data class-action exposure modeled at $750-$150,000 per work under 17 U.S.C. § 504.

Should we hire Latham, Cooley, Wilson Sonsini, or Kirkland for IP DD on an AI target?

The AmLaw 100 IP DD market splits into four positioning tiers for AI-target work in 2026. (1) Latham & Watkins — broadest cross-practice integration with the M&A team, fastest at running parallel patent + trademark + AI litigation diligence, strongest for large-cap PE buyers ($500M+ EV). (2) Cooley — deepest sector-native AI/ML practice from a quarter-century of representing AI-native targets, strongest for AI-specific training-data and model-weight diligence, fast IC-memo turnaround. (3) Wilson Sonsini — strongest patent prosecution + Silicon Valley founder-network credibility, ideal when target is a Stanford/Berkeley-rooted spinout with patent-heavy R&D. (4) Kirkland & Ellis — bulge-bracket PE M&A integration, strongest when IP DD must fold seamlessly into a multi-billion-dollar LBO and the R&W carrier wants Kirkland-branded reps. Skadden, Cravath, and Wachtell each have credible IP DD practices but are usually retained for the M&A side, not the IP workstream specifically. The 5-Asset Encumbrance Matrix is methodology-agnostic — pick the firm for the practice fit, not the framework. Boutique IP firms (Fenwick, Morrison Foerster, Knobbe Martens, Finnegan, Sterne Kessler) compete on patent-heavy targets but rarely run the full 5-asset audit on M&A timelines.

What are the 7 IP Killers that have re-priced or killed deals in 2025-2026?

The 7 IP Killers framework — derived from 2024-2026 deal-anchor analysis — covers (1) Defective Patent Assignment Chains (AIA assignee-applicant gaps; routinely 4-5 on the title-chain axis), (2) OSS License Contamination (AGPL/SSPL/BSL in core product; remediation cost $200K-$2M), (3) AI Training-Data Liability (Bartz v Anthropic $1.5B settlement preliminarily approved Sep 25 2025; UMG v Anthropic $3B+ second lawsuit filed Jan 28 2026; NYT v OpenAI 20M log production order affirmed Jan 5 2026), (4) Trademark Non-Use and Foreign Filing Gaps (50-jurisdiction portfolios with 18-month non-use windows), (5) Trade-Secret Disclosure Without Protection (Appian v Pegasystems $2.036B trade-secret verdict vacated and remanded for new trial by Virginia Court of Appeals July 2024, affirmed by Virginia Supreme Court January 2026; Waymo v Uber $245M settlement Feb 2018), (6) Open Litigation IP Overhang (Disney + Universal v Midjourney filed June 11 2025; Getty v Stability AI UK High Court ruling Nov 4 2025), (7) R&W Underwriter Exclusion Cascade (AI training-data, pending IP litigation, and patent invalidity claims excluded from coverage by default). Each of the 7 IP Killers maps to a specific cell in the 5-Asset Encumbrance Matrix and a specific score band. Killers #1, #2, and #3 are the three most common 71+ score drivers in 2026 software and AI M&A.

Related resources

• IT due diligence (2026) — the 6-Axis Tech-Stack Fragility Audit; license stack axis covers IP transferability at the SaaS contract level
• Environmental due diligence (2026) — the BFPP Defense Stack + PFAS Exposure Surface Map for industrial and real estate targets
• M&A due diligence process guide — the 6-phase playbook the 5-Asset Encumbrance Matrix sits inside
• Operational due diligence — the 8-System ODD Audit + 3-Axis Severity Matrix (IP DD feeds the Legal Operations cell)
• AI due diligence — the 5-Layer AI Target Audit (Data Layer covers training-data IP at the AI-stack level)
• Due diligence questionnaire — the 5-Persona DDQ Template Library (IP-specific subset = 40-80 questions)
• Due diligence data room checklist — the file-side companion for staging the 5-Asset folder structure
• SaaS M&A data room 2026 — sector-specific application of the 5-Asset Encumbrance Matrix for SaaS targets
• Peony data rooms — the IP-DD-grade data room engine
• Peony visitor groups — workstream isolation across IP counsel, R&W underwriter, deal team
• Peony page-level analytics — leading indicator of which 5-Asset Encumbrance Matrix axis will score red
• Peony watermarks + screenshot protection + leak protection — DTSA reasonable-measures controls for trade-secret asset class
• Peony NDA gate + link expiry + manage links — disclosure schedule cascade controls
• Peony AI auto-indexing — 5-asset folder structure auto-built in under 3 minutes
• Peony smart Q&A — buyer IP counsel questions routed to seller IP counsel with audit trail
• Peony redaction + manage links — disclosure schedule subroom controls
• Solutions: M&A | Private Equity | Legal | Technology Sales | Biotech | Media

---

Deqian Jia is co-founder of Peony, the data room engine PE deal teams and outside IP counsel use to stage their IP DD evidence pack. The 5-Asset Encumbrance Matrix and 7 IP Killers frameworks are Peony's contributions to the 2026 IP DD practitioner playbook — built from 2024-2026 deal anchor analysis across software, AI/ML, biotech, deep-tech, and consumer M&A.