ClaimHit / Confidentiality
Your patents are your business. Not our training data.
A search tells us which patent you are interested in and which companies you think might be reading on it. That is sensitive before it is anything else. Here is exactly what happens to it.
Five questions, answered plainly
Do you train on our work?
No. ClaimHit trains no model and fine-tunes no model. There is no training pipeline in the product.
The analysis runs on third-party models called through their business APIs, under terms that do not permit your inputs to be used for training. We do not sell, licence or share your searches with anyone.
Do you keep our results?
Yes, in your account — the point of a Hit Chart is opening it again next quarter.
They stay until you delete them. Nothing expires on its own, and nothing is archived somewhere you cannot reach.
What happens to documents we upload?
The raw file is deleted within 24 hours, and you get an email confirming which files went.
A teardown report or internal spec is read once and the relevant evidence is folded into your chart. That extracted evidence stays, because it is the thing you paid for. The file itself does not.
What happens when we delete the account?
Your searches and the charts in them are deleted — not de-identified and kept. So are your profile, transaction records and team membership.
One exception: expert-review orders are anonymised and held for seven years, because they are commercial records of work performed. That is a legal retention duty, and it is the only one.
Who at ClaimHit can see it?
Support and engineering staff can reach account records where a problem requires it, under access controls and logging.
Nobody browses customer searches, and we do not use the content of your work to prospect — in your market or anyone else’s.
At a glance
- Training on your data. Never, by us or by a provider.
- Uploaded files. Deleted within 24 hours, with a confirmation email.
- Sold or shared. No, to either.
- Searches and charts. Yours until you delete them.
- Account deletion. Searches and charts deleted outright.
- Expert-review orders. Anonymised, kept 7 years, as commercial records.
Where your text actually goes
A product that calls a model cannot tell you nothing leaves. What it can tell you is what leaves, to whom, and under what terms — which is the claim that survives your own security review.
To the models
The patent, the claim and the evidence pass through third-party model providers, because that is what performs the analysis.
They are called through business APIs, where inputs are not used to train models, and where a provider offers zero data retention we take it. The providers are named on request, under NDA if you prefer.
To web search
Finding who implements a claim means searching the open web, so search vendors are involved. What they receive is generated search queries — company names, product categories, technical vocabulary — not your claim text.
Those queries are derived from the claim, so we will not promise that no claim-shaped language can ever appear in one. We will tell you the claim itself is not what is sent.
No patent number and nothing identifying you goes to a search vendor. Not your name, your company, your account, nor any per-user identifier their API offers. The request carries the query and nothing else; our own key is what identifies the caller, and it identifies ClaimHit, not you.
That is deliberate, because a search vendor may log the queries it is sent and we do not control their retention. So the protection does not rest on what they keep. It rests on a kept query being a phrase about a product category, with nothing in it that leads back to you or to which patent was being read.
To the services that read a page
Some vendors publish documentation their own site will not hand to a plain fetch — a page built entirely in the browser, or a server that refuses us. Reading those needs a service that loads the page properly, so one is used where a direct read comes back empty.
What it receives is a web address, and nothing else — a public page we want the text of. Not your patent, not the claim, not your identity. A list of URLs does say which products were looked at, which is why those services are named in our subprocessor list on request like every other one.
Unpublished and draft claims
ClaimHit accepts claim text directly, so a claim being drafted or amended can be assessed before any number exists to look it up by.
That capability deserves its warning: an unpublished claim processed this way still passes through the model providers above. If your matter requires that a draft claim touch no third party at all, have that conversation with us before you run it, not after.
What we never ask for, and never share
- Your client list. We do not ask who the patent is for.
- Your strategy. Nothing in the product asks why you are looking, and nothing records an answer.
- Your results, to anyone else. Not sold, not licensed, not pooled into a dataset.
Ask us anything specific
Security reviews, DPAs, subprocessor lists and NDAs are ordinary requests here and a person answers them. Tell us what your firm needs to sign off.
What comes back is research, not a determination
Every assessment and every chart carries a disclaimer saying so. Results are candidates worth investigating, cross-checked against public evidence you can open and read. They are not a finding of infringement, essentiality or validity, and a qualified practitioner has to review them before anyone relies on them.
That is a confidentiality point as much as a legal one: what you are holding is a research product, and the privilege position of what you do with it is yours to manage with your own counsel.
See who is using your patent.
The pilot is one patent: the full product search and one Hit Chart, with a 30-day money-back guarantee. No subscription, no card on file after that.
One patent, the full search and one Hit Chart. 30-day money-back guarantee.