What if your records contain someone else's copyrighted material?

Does owning your records mean you own everything inside them?#

Owning your records does not mean owning every item inside them. The company generally owns what its employees write within the scope of their jobs, because the Copyright Act treats that work as made for hire, with the employer considered the author. Material created by others keeps its own owner, even when a copy lives in your Zendesk, SharePoint or Procore account.

The distinction matters once records leave the company under a license. A work order note written by your technician is yours; the manufacturer's service bulletin attached to it is not. A license that delivers both, without permission for the bulletin, may put the supplier in breach of its own rights warranty and may expose it to a claim from the bulletin's owner.

Contractors are a separate case. The Copyright Office explains that work by a non-employee is made for hire only in certain listed categories and only with a signed written agreement, so documents written by freelancers usually need an assignment before the company can treat them as its own.

Where third-party material hides in business records#

Third-party material usually hides in attachments, quoted text and files received from clients or vendors. It is rarely the bulk of an operational archive, but it turns up in nearly every system and is easy to miss in a quick review.

Why buyers now ask about provenance#

AI buyers ask about provenance because recent copyright decisions turned on where training material came from and how it was used. In Bartz v. Anthropic, a federal judge held in June 2025 that training on lawfully purchased books was fair use, but that downloading and keeping pirated copies in a central library was not excused.

In Thomson Reuters v. Ross, a federal court in Delaware rejected a fair-use defense in February 2025 for copying Westlaw headnotes to build a competing AI legal-research tool, and the Third Circuit was reported in late September 2026 to have affirmed. In Kadrey v. Meta, the court found fair use on the record before it but stressed that it was not holding AI training lawful in general.

These cases concern AI developers' own conduct, not companies that license records to them. Their effect on suppliers is indirect but real: buyers want every delivered item to come with a clear rights story. The US Copyright Office's May 2025 pre-publication report on generative AI training also addresses licensing copyrighted works for training, which keeps rights documentation at the center of these deals.

Exclude, strip or clear: how to decide#

The choice between excluding, stripping and clearing depends on how central the third-party material is to the record and whether its owner can be reached. Stripping works when the outside item is separable from work your staff wrote. Exclusion fits record families where outside material is the substance. Clearance makes sense when the item matters and one identifiable owner can grant permission.

A first sweep can look for signals rather than reading every file: attachment file types, copyright notices, phrases such as all rights reserved or licensed to, identical PDFs repeated across many tickets, and sender domains of publishers and equipment makers. Reviewers then sample what the sweep flags and confirm the rule for each record family.

• 1Exclude when the record family is mostly someone else's work, such as client deliverables, received drawing sets or purchased course libraries.
• 2Strip when the outside item is separable, such as an attached PDF manual or a pasted article, and the surrounding discussion stands on its own.
• 3Reference instead of reproducing when staff cite a standard or manual: keep the citation and the reasoning, remove the copied text.
• 4Clear when a single owner, such as a client, can give written permission and the material adds real value to the package.
• 5Log every decision with the record family, the rule applied, the reviewer and the approver.

Contract terms that settle the question#

Contract terms on both sides settle much of the answer. Inbound agreements with clients, vendors and design professionals may already say who owns what you received, while the outbound data license sets what you promise the buyer about rights.

Construction records show how inbound terms work. Under the AIA A201-2017 general conditions, the architect and its consultants are deemed the authors and owners of their instruments of service, and contractors may use and reproduce them only for performing the work. A contractor should assume it does not own drawings it received, whatever system now holds them.

Illustrative: a mechanical contractor's service archive#

Illustrative: a fictional mechanical contractor runs service work in ServiceTitan and construction projects in Procore. Its service records include technician notes, photos and, on many work orders, attached manufacturer installation manuals and warranty bulletins. Its project records include RFIs, submittals and the architect's drawing sets.

Outside counsel and the operations lead agree on three rules. Manufacturer PDFs are stripped from work orders, while technician notes, model numbers and fault descriptions stay. Drawing sets and submittal attachments are excluded entirely. RFI questions written by the contractor's own project managers are kept, and design-team responses are held back pending a closer rights review.

The licensable package is smaller than the raw archive, but every item in it has a known owner. The exclusion log goes into the deal file, so the buyer's reviewers can see what was removed and why.

How SourceX handles embedded third-party content#

SourceX handles embedded third-party material in the Rights and Preparation steps of the SourceX five-step transaction: Supply, Rights, Preparation, Approval and Delivery. The rights review identifies record families that carry outside material, preparation strips or excludes it alongside personal and confidential details, and the supplier approves the final scope before anything is delivered.

The decisions are recorded in the SourceX Evidence Packet, which covers provenance, licensing rights, permitted use, the privacy record and release authorization. Which laws and contract terms apply is assessed deal by deal with counsel, and both sides review one record of what was included, what was removed and on what basis.