Is Free Online OCR Safe for Legal Documents?
By Schema LawPublished Updated
Free online OCR is safe for legal documents that are already public or contain nothing confidential. It is not a safe default for client documents, because a free converter's terms may keep the right to store, review or reuse what you upload. Read the terms first, and move client files to a tool with written commitments.
The price is not the problem. Some free tools treat files carefully and some paid ones do not. What matters is whether you can find out what happens to an upload, and whether you could explain the answer to the client. This post is not legal advice, and confidentiality rules vary by jurisdiction.
How do free OCR sites pay for themselves?
Free OCR sites pay for themselves in one of four ways: advertising, selling upgrades, selling the same technology to businesses, or making use of what people upload. The first three need your attention or your later custom. Only the fourth needs your document, so work out which model you are looking at.
| How the site earns | Signs you will see | What it means for your file |
|---|---|---|
| Advertising | Banner ads around the upload box, many third-party trackers | The file may be left alone, but the operator has little reason to invest in protecting it |
| Upgrades | Page limits, a pricing page, prompts to subscribe | The free tier is a sample. Terms may match the paid plans or may be weaker |
| Business sales | Developer documentation, a "contact sales" link | The free page is a demonstration. Business customers may get terms the free page does not |
| Use of uploads | Terms that mention improving services, research, or a broad licence to your content | Your document is part of what the operator gets in return |
A site can combine several of these. The fourth is the hardest to spot from the page itself, which is why the terms matter more than the design.
Bar authorities have noticed the pattern. The D.C. Bar's Ethics Opinion 388 on generative AI in client matters describes a "free" service as one "that the provider makes available for the provider's own marketing and product development purposes", and says business users who pay may be able to negotiate better terms. The opinion is about generative AI and speaks for the District of Columbia, but the economics it describes are the same for a free converter.
What do free converters say about your files?
Free converters' terms can address five things about your files: the licence you grant, how long uploads are kept, whether they are used to improve the product, who else receives them, and what happens if the business is sold. Some sites answer all five clearly. Some answer none.
You will not find these under a heading called "confidentiality". Look for sections titled "Your content", "User content", "Licence", "How we use information", "Sharing" and "Retention".
| Wording to look for | What it means | Acceptable for a client file? |
|---|---|---|
| "Files are deleted automatically after [a stated period]" | A fixed retention period. Check whether backups are included | A good sign, if the rest holds up |
| "You grant us a licence to use, reproduce, modify and distribute your content" | The operator may do more than convert the file. Read what the licence is limited to | Not without a limit to providing the service to you |
| "We may use content to improve our services" or "for research" | Uploads may be studied or used to train software | No |
| "We share information with service providers, affiliates and partners" | Other companies receive data. Check whether file contents are included | Only if those recipients are bound to the same limits |
| "Information may be transferred in a merger, sale or bankruptcy" | A future owner inherits what is stored | Depends on how long files are kept |
| "We may change these terms at any time" | Today's promise can be withdrawn | Save a dated copy and recheck |
| Nothing about files at all | You have no commitment to rely on | No |
The D.C. opinion shows how such clauses read in the wild. It quotes one provider's consumer privacy policy, which said the company may use what users submit "[t]o improve our Services and conduct research" and "[t]o develop new programs and services", and which listed "Business Transfers" among the occasions for disclosure. The committee's conclusion was that lawyers who share client confidential information with providers under policies like that risk violating their confidentiality obligations, absent client consent.
One more point catches people out. The State Bar of California's Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law notes that "a user may agree to the terms of use and privacy policy simply by using the product." On a converter with no sign-up, dropping the file onto the page can be the acceptance.
Read what is missing as carefully as what is there. A privacy policy that talks only about cookies and email addresses is describing visitors to the website. It may say nothing about the files, and silence is not a promise.
Vendor security on a no-account page is often a blank you should treat as a stop. Third-party risk shows up in trackers and in any company the wording lets the operator share files with. Client confidentiality is difficult to explain if you cannot name who received the upload. The duty of competence is the reason those two minutes come before a client page.
Is any free OCR acceptable for client documents?
Yes, in limited cases. Free OCR can be acceptable for client documents when it runs on your own computer, when a free tier carries the same written commitments as the paid plans, or when the client has given informed consent to the specific risk. Being free does not disqualify a tool. Unknown terms do.
The three cases in more detail:
- Free software that runs locally. An open source engine installed on a firm computer sends the document nowhere. The risks are those of the device. The post on private OCR compares local tools with hosted ones.
- A free tier with real terms. If the provider's commitments on access, reuse, retention and deletion apply to free accounts as written, the free tier meets the same test as the paid plan. Confirm that the terms say so, because some providers reserve their stronger terms for paying customers.
- Informed client consent. The California guidance says a lawyer "must not input any confidential information of the client into a generative AI solution that may present material risks to confidentiality or security, absent informed client consent". Consent has to be informed about the actual risk, which is hard to do when you cannot tell what the site does with a file. The D.C. opinion adds that clients "typically should not be asked to consent" to wide-ranging disclosures.
Both documents address generative AI, and each binds or guides lawyers in one jurisdiction. Whether and how your bar applies the same reasoning to a file converter is something to check locally.
A common mistake is to reason from the size of the risk to its acceptability: "It is only one page, and nobody at that site will ever look." The duty concerns what you made possible, and you cannot know who looks. A second mistake is to treat removal of the client's name as enough. A lease, a medical chart or a bank statement identifies its subject in many places besides the first line.
Documents you can run through a free tool
You can run a document through a free tool when its contents are already public, or belong to nobody but you, and when the upload itself tells an observer nothing about a client's matter. That covers more daily work than people expect, and it keeps the careful route clear for the files that need it.
Usually fine:
- Published court opinions, statutes, regulations and agency guidance that will not copy cleanly
- Blank court forms and government forms
- Pleadings and orders on a public docket that are not sealed or restricted
- Your own firm's published articles, newsletters and marketing material
- Textbook or treatise pages you are entitled to copy, for your own notes
Keep out of a free tool with unknown terms:
- Anything a client sent you or you prepared for a client
- Productions from another party, especially under a protective order
- Medical, financial, employment and immigration records
- Drafts, notes and internal memoranda
- Sealed or redacted filings, including the unredacted version of a public one
One edge case deserves a pause. A public document can still be confidential in your hands. In some jurisdictions the duty reaches well past privileged material. The D.C. opinion, for example, explains that the D.C. rule protects both privileged "confidences" and a wider class of "secrets" gained in the professional relationship. Check how your own rule is drawn before treating "public record" as the end of the question.
Free tiers of paid services versus anonymous converters
A free tier belongs to a company that also sells the service, wants you as a customer and usually publishes terms that cover your files. An anonymous converter is a page with an upload box and no account, where you often cannot tell who operates it. The two are both free and have little else in common.
| Free tier of a paid service | Anonymous converter | |
|---|---|---|
| Who runs it | A named company with a pricing page and contact details | Often unclear |
| Account | Required, so uploads are tied to you and can be deleted by you | None, so you cannot return to delete a file |
| Terms about files | Usually present. Check whether free accounts are treated differently | Sometimes absent |
| Why it is free | To let you try the product | Advertising, or unknown |
| Limits | Page or document caps | Often file-size caps, sometimes none stated |
An account can feel like a privacy cost, since you give an email address. For client files it is a benefit, because it gives you a place to see what you uploaded and remove it.
Do not assume a free tier inherits everything the paid plans promise. Look for a sentence that says the commitments apply to all plans, or ask.
Schema Law's Free plan is a free tier of this kind. The pricing page lists it as one document a day, up to three pages per document. The same page lists six lines under a heading that applies to every plan, Free included. Four concern how documents are handled: "Your documents are never used to train an AI model", "Documents stay inside your own account", "Encrypted in transit and at rest" and "Delete any document, at any time". The other two give every plan all eight export formats and uploads of PDF, PNG or JPEG files of up to 20 MB. Nothing is deleted automatically, so a document stays in the account until you delete it. Those are the company's own statements, and the check in the next section applies to them as it does to anyone's. The limits of free tiers in general, and when three pages is enough, are covered in the post on free scanned PDF to Word conversion.
A two-minute safety check for any free converter
Before uploading anything that is not public, spend two minutes on six checks: who runs the site, whether the connection is secure, what the terms say about files, whether uploads are reused, how long they are kept, and whether you can delete them. Any answer you cannot find counts as a no.
- Who operates it? Scroll to the footer. Look for a company name, an address or country, and a contact. No operator, no upload.
- Is the connection encrypted? The address should begin with
https. That protects the file on its way there and nothing more, as the post on encrypted OCR explains. - Do the terms mention uploaded files at all? Open the terms and the privacy policy and search each for "content", "files" and "upload".
- Is content reused? Search for "improve", "research", "train" and "licence". Any clause that lets the operator use file contents for its own purposes ends the check.
- How long are files kept? Search for "retain", "delete" and "days". Look for a stated period and for whether you can delete sooner.
- Can you delete, and prove it? If there is no account, you cannot. If there is, find the delete control before you upload.
Score it simply. Six clear answers in your favour means the tool may be a candidate for routine client documents, subject to your firm's own approval. Any gap means public documents only. A reuse clause, or no identifiable operator, means do not use it for firm work.
Two minutes is a screen, not a review. A tool that passes and that the firm wants to rely on deserves the fuller set of questions in the post on whether it is safe to upload client documents to an OCR tool, and a line in the firm's register of approved tools.
If you have already uploaded a client document to a site that fails the check, tell the lawyer responsible for the matter the same day. The guide to a law firm AI policy for document uploads sets out the steps to take.