Updated October 2026
Selling IP and patent prosecution firm data for AI training: what it's worth and who buys it
Invisible says it draws reinforcement learning tasks from legal and compliance work, and patent prosecution is legal reasoning written at length against a technical record.1
Patent prosecution produces some of the longest structured reasoning any law firm writes. An invention disclosure becomes claims and a specification; an examiner rejects some claims over prior art; the firm answers with amendments and arguments; the cycle repeats until allowance or appeal. Docketing tracks every statutory deadline, and annuity reminders run for decades after grant.
Buyers describe wanting exactly this kind of record of skilled, multi-step work. micro1 names legal and compliance and lists templates, playbooks and task histories. Invisible says its reinforcement learning tasks come from legal and compliance work, and Mercor names legal among its eight industries. Paralegals and legal secretaries, who run docketing, score 100 out of 100 on O*NET's email scale.
Patent files carry their own sensitivities: unpublished applications, client invention disclosures, and technology that may be export-controlled. The part of a prosecution archive to describe first is the firm's method, the part least tied to any one client: claim-drafting guides, office-action response templates, the docketing rulebook and the formalities checklists. Raise everything that describes a client's invention, published or not, with the licensing company before any export.
Worth
What a firm's data may be worth
Pre-set for a typical firm: 35 staff and 15 years of records. No buyer publishes pricing by industry, so industry does not move the band. Our estimate, not a buyer quote.
Buyers
Which buyers want this data
Each reason is the buyer's own published statement about the tools, data or industries it takes, matched to how IP and patent prosecution firms work.
Featured partners
micro1
Names legal and compliance and lists templates, playbooks and task histories, the form a firm's claim-drafting guide and docket history take.10,11
- $100K to $2M+, micro1 says7
Watch out for: Says it removes personal details only "when relevant", so ask exactly how.12
Everyone else we reviewed
By score, with full reviews. We have no referral agreement with these companies, so we do not link out to them.
Mercor
Names legal and connects to Outlook, Teams, Drive and Confluence, where prosecution teams usually keep drafting guides and examiner-interview notes.13,14
Watch out for: It confirmed a cyberattack in 2026, and class actions followed.18,19
Troveo
Licenses on a revenue share rather than buying outright and starts with a free value assessment, which suits a firm that wants to test interest before scoping.20,21
- Six figures and up for a whole company, Troveo says22
- A share of the revenue; the percentage is not published20
Watch out for: You are paid from a revenue share, so ask what percentage you keep.20
Buyer demand
How buyers' own tools rate IP and patent prosecution firms
No buyer publishes a price list by industry. These are settings inside buyers' and brokers' own estimators and intake forms, read from their public pages on 2026-10-14. They show how each company sorts industries, not what any firm was paid. The category shown is the one a firm would most likely pick; that match is ours.
| Company | Its category | How it rates the industry |
|---|---|---|
| Troveo | Legal | Tier 2 of 4: "Your industry carries above-average buyer demand." Price factor 1.15 in its estimator.23 |
| Polyshares | Legal | Marked premium in its intake list24 |
| VMS | Law | Higher weight in its estimator (factor 1.30)25 |
| Ventra | Legal, healthcare, financial services or accounting | Says it cannot take these industries right now26 |
Size rules buyers publish
Checked against 35 staff, the typical firm size this page's calculator is set to. Ask each buyer whether contractors and part-time staff count.
Polyshares
No headcount rule; the length and texture of the record matter more27
Mercor
None stated28
Troveo
No minimum stated; its estimator starts at 2029
Nyne
20+ US W-2 employees and 3+ years30
Handshake AI
20+ full-time employees and 3+ years31
micro1
Pewter
20+ staff and 3+ years33
Endless
Roughly 30+ employees34
Ventra
30+ US employees and 3+ years of records35
Supersoniq
50+ employees and 5+ years36
What buyers value
What in a firm's records buyers value
The column on the right is not legal advice: it lists what to raise with the buyer before anything is exported.
| Data | Where it lives | Why a buyer values it | Raise with the buyer |
|---|---|---|---|
| Claim-drafting and specification guides | KM wiki or shared drive | How the firm turns a disclosure into claims of different scope, written down by senior practitioners. | Nothing specific |
| Office-action response templates | Word templates and precedent folders | Argument structures for common rejection types, refined over thousands of responses. | Templates that still carry a client's claim language or technology. |
| Docketing rulebook and deadline history | Anaqua, CPI or AppColl | The rules for computing and reminding every statutory and foreign deadline, plus a record of how each was met. | Application numbers and client names attached to each docket entry. |
| Formalities and filing checklists | Shared drive | Paralegal procedures for filings, assignments, information disclosure statements and foreign filing. | Nothing specific |
| Examiner interview notes and strategy threads | Outlook, Teams | How attorneys plan and debrief an interview, and what arguments moved an examiner. | Unpublished application content and client strategy in each thread. |
| Invention disclosure intake forms | Client portal or web form | The questions the firm asks inventors to get a draftable disclosure, a reusable intake design. | Completed disclosures, which describe client inventions. |
A week of records
A week of data at a firm
- Monday: the docket report lists 46 deadlines for the fortnight; two are reassigned in Teams.
- Tuesday: a non-final rejection arrives; the associate starts from the firm's template for that rejection type.
- Wednesday: an examiner interview is held and the agent's debrief note goes to the partner.
- Thursday: a senior partner adds a section on means-plus-function claims to the drafting guide.
- Friday: formalities staff file four continuations from the checklist and update the docket.
Process
How it works with each buyer
In each buyer's own published words. Where a buyer publishes nothing, we say so rather than guess.
micro1
- Internal docs, SOPs and playbooks, knowledge bases, project management workflows, internal comms, templates, CRM and workflow metadata, task histories11
- 30 to 200 employees32
- $100K-$2M+ per data package; over $1M for unique data7
- PII removed "when relevant"12
- The company keeps ownership of its data and reviews samples before release37
- Not published
Mercor
Selling or closing
Selling or closing your firm? License its data first
IP boutiques are regularly absorbed by larger firms, and when that happens the drafting guides and docketing rules often go with the people while the archive stays behind. Before a merger or sale, ask each licensing company whether the firm's method material can be licensed separately from client files. Buyers already approach companies around deals: Mercor went to Warmly about its archives after HubSpot acquired it.2
FAQ
IP and patent prosecution firms: questions, answered
The method: drafting guides, response templates, the docketing rulebook and formalities checklists. Endless lists templates, precedents and process documentation as in scope for law firms and keeps client matter content out. Client claim language and disclosures are the part to raise with the buyer.40
The deadline rules and task history are workflow records of the kind micro1 lists as task histories and project management workflows. Application numbers and client names attached to each entry are what to discuss with the buyer before export.11
No buyer publishes patent-firm pricing. Published ranges mostly run $100K to $2M, as the buyers' own claims: micro1 states $100K-$2M+, Cognor Labs $200K-$2M on its partners page, and Jottings $1M+ for rare data. A deep drafting guide and long docket history are the parts that argue for the top of those ranges.7,43,44
We do not answer legal questions. Ask the licensing company: their team reviews this with you before anything is signed.
We do not answer legal questions. Ask the licensing company: their team reviews this with you before anything is signed.
We do not answer legal questions. Ask the licensing company: their team reviews this with you before anything is signed.
We do not answer legal questions. Ask the licensing company: their team reviews this with you before anything is signed.
Before you sign
Legal questions go to the licensing company
We do not answer legal questions. Ask the licensing company before you sign: their team reviews these with you.
Questions for this kind of business
- How will unpublished applications and invention disclosures be kept out or handled?
- Does your review check for export-controlled technical content (EAR, ITAR)?
- Do client engagement terms on confidentiality of inventions limit what can be included?
- Who owns this data: the company, an estate, or someone else?
- Do customer, vendor or platform contracts, NDAs or API terms restrict this?
- Could the data include privileged material or trade secrets?
- Do professional rules apply (AICPA, IRC 7216, HIPAA, ABA Model Rule 1.17, insurance agency contracts, adviser rules, state boards)?
Questions for any buyer
- Is the license exclusive, and for how long?
- Can you sublicense the data, and to whom?
- Which AI labs will receive it, and will you tell us?
- How long do you keep it, and how is it deleted?
- What is your liability cap, and who indemnifies whom?
- Will the contract ban re-identification of people in the data?
- Who removes personal data, by what method, and do we approve it first?