Rights and privacy

Public data, scraping, and fair use: separate questions

Avoid treating public access, collection method, and downstream licensing authority as interchangeable claims.

Sources and guidance · Reviewed October 2026

Start with the exact conduct

Publicly viewable content, permission to access a website, permission to copy material, and authority to license a dataset are distinct questions. Write down how the material was obtained, what was copied, which terms or restrictions applied, and what you now propose to grant a buyer.

Preserve collection dates, source URLs, relevant terms, and component provenance. If an upstream supplier collected the material, ask for its evidence. Do not convert an uncertain collection history into a confident statement that the dataset is cleared because a search engine could find the original pages.

Do not turn fair use into a blanket warranty

The U.S. Copyright Office explains that fair use involves a fact-specific assessment of statutory factors, including purpose, the work, the amount used, and market effects. A general claim that AI training is research does not settle every use or jurisdiction.

Have qualified counsel evaluate the actual activity and applicable law. Avoid presenting a court decision about one set of facts as a universal license for unrelated collection, redistribution, or commercial training. The seller’s proposed sublicense is a separate commercial commitment that deserves careful review.

Check additional constraints

Copyright is not the only issue. Material may contain personal information, confidential content, trademarks, or contractually restricted components. The collection method may raise separate access or contractual questions. A technical crawler setting does not answer all of these.

Keep access controls and security boundaries intact. Do not bypass a restriction merely to assemble a larger collection. If the rights basis is uncertain, segregate the source and consider a permission-based route or a different dataset. Document the limitation instead of hiding it in a generic disclaimer.

  • Separate original content from quoted or embedded material.
  • Preserve source and collection-method evidence.
  • Record unresolved legal questions and excluded sources.

Make buyer discussions accurate

Tell the buyer which rights you claim and the basis for those claims. If the proposal depends on a legal interpretation rather than an express license, make that distinction clear to advisers and counterparties. Do not promise exclusive ownership of material you do not own.

A narrower collection of original or expressly licensed content may be easier to diligence than a broad scraped archive. That is a commercial option, not a conclusion that every scraped dataset is unlawful. The right next step is a transaction-specific review, followed by an honest decision about what can be offered.

Common questions

Does robots.txt grant a data license?

Do not treat crawler instructions as a complete rights clearance. Review access conditions, copyright, privacy, and the proposed downstream use separately.

Can a buyer accept the legal risk for me?

Risk allocation can be negotiated, but a contract cannot manufacture rights that do not exist. Get advice on the actual claims and obligations.

Sources & further reading

U.S. Copyright Office: More information on fair use ↗
Use this as a preparation framework. It is not legal advice, a valuation or a promise that a buyer will accept your data.