This post is general information, not legal advice. The law on AI and intellectual property is unsettled and changing, it differs by country, province and state, and facts matter. Nothing here is a statement of what the law is for your situation. We are not lawyers writing to you as a client. Talk to a lawyer licensed where you operate before you rely on any of it.
With that said, most business owners do not need a legal treatise. They need to know which questions to ask. This post sets out five, as of November 2026, with the public sources we could verify, and says plainly where we could not settle a point.
1. Who owns what AI makes?
In the United States, copyright protects human authorship. The U.S. Copyright Office's January 2025 report, *Copyright and Artificial Intelligence, Part 2: Copyrightability*, concluded that prompts alone do not make a user the author of what an AI produces, while using AI as a tool does not by itself remove protection for the human parts of a work: for example, original material you add, or your selection and arrangement, or modifications you make. The Office said these questions can be handled under existing law.
A federal appeals court reached a related result in *Thaler v. Perlmutter* (D.C. Circuit, March 18, 2025). It held that the Copyright Act requires a human author, and affirmed the refusal to register an image that the claimant said was made autonomously by an AI system. The court left for another day how much human contribution is enough.
In Canada, we did not find a settled rule for AI-generated works. The federal government consulted on this from October 2023 to January 2024, covering training, authorship and ownership of AI-generated content, and liability, and published a "What We Heard" report in February 2025. As summarized in coverage of that report, stakeholders were divided on whether existing law is enough. A consultation report is not law.
What it means in practice. Do not assume you hold exclusive rights to a logo, an image or a block of text that a model produced from a short prompt. If exclusivity matters, for example for a brand mark or a catalogue photo style, involve a human creator, keep records of the human contribution, and ask a lawyer how to protect it. For routine content such as an email or a product description, exclusive ownership may matter less than accuracy.
2. Training data: can a model's makers be sued, and does it touch you?
Many AI models were trained on large amounts of text, images and code collected from the internet and elsewhere, some of it protected by copyright. Whether and when that is lawful is being argued in courts and policy bodies in several countries. In the U.S., the Copyright Office published a pre-publication Part 3 on generative AI training in May 2025, and there are many lawsuits. We are not going to summarize the cases here: positions differ, they move, and some are on appeal.
Two things to take from it without taking a side:
- It is unsettled. Anyone who tells you it is definitely fine, or definitely not, is simplifying.
- It can reach you indirectly. If a vendor's output closely copies protected work, a business that publishes it could face a claim. That is why indemnity terms (below) matter.
3. Confidentiality: what you paste in
This is the risk most businesses meet first, and it does not depend on any unsettled law. When you paste a contract, a customer list or a draft into an AI tool, where does it go, who can see it, and is it used to train anything?
Ask for the answers in writing:
- Are your inputs and outputs used to train the provider's models? Is there a setting or contract term that switches that off, and is it on by default?
- How long are they retained, and can you delete them?
- Which sub-processors handle them, and in which countries?
- Who at the provider can read them?
Trade secrets can lose their protection if they are not kept confidential. Pasting something sensitive into a consumer tool whose terms allow broad use of inputs is the kind of thing a lawyer will ask about later.
Personal information. If you handle personal information about customers or staff, privacy law applies. In Québec, the private-sector privacy act, as amended by Law 25, includes obligations on consent, on transfers outside Québec and on automated decisions. Section 12.1 covers decisions made exclusively by automated processing, and it requires informing the person. You can read the Act on Légis Québec. In Canada's other provinces and federally, PIPEDA or provincial laws apply; in the US, a patchwork of state and sector laws applies. Whether and how any of these apply depends on your facts, so ask your privacy lead or lawyer, and read the text itself.
4. Contracts and terms to check
When you buy or sign up for an AI service, read these parts of the terms. Better, give this list to your lawyer.
- Ownership of outputs. Does the provider assign you the rights it has in the outputs, or only license them? Are there restrictions?
- Use of your data for training. Default, opt-out, or never. Get it in the contract, not only the help centre.
- Confidentiality and security. Obligations, certifications you can verify, incident notice.
- Data location and sub-processors. Where processed and stored; which sub-processors; how you are told about changes.
- Retention and deletion, including at the end of the contract.
- IP indemnity. Some providers offer to defend customers against certain IP claims about outputs, with conditions. Read the conditions: what is excluded, what you must do, any caps.
- Warranties and liability limits. Many AI terms disclaim accuracy and cap liability. Know what you carry yourself.
- Your own customer contracts. Do they restrict using third-party AI on their data? Do you need to tell them?
- Employees and contractors. Do your agreements and policies say what staff may paste into AI tools, and who owns work made with them? Update them.
- Language and form. If you contract in Québec, ask your lawyer about language requirements for contracts and notices.
5. A few habits that help
- Keep a record. For anything you may want to own or defend, such as a logo, a manual or a design, note who did what, and keep versions.
- Label AI-assisted material internally, so you know which parts have a human author.
- Write a one-page AI use policy. What tools are approved, what data may go in, who approves publishing.
- Have a person approve what goes out. A person reading an AI draft before it is sent or published is a basic control for accuracy, privacy and copying alike.
What this means for your business
- Treat AI output as useful but not automatically yours exclusively.
- Treat training-data lawsuits as unsettled, and look for clear contract terms rather than reassurance.
- Treat confidentiality as the practical risk today: know where your data goes.
- Read the terms on ownership, training use, retention, indemnity and sub-processors.
- Get legal advice for anything important: brand assets, regulated data, big contracts.
How this shows up at Humonix
Our clients' data is theirs, and we give a data processing agreement and a list of our sub-processors; see our DPA, subprocessors and privacy policy, and Hue's security page. Those documents, not this post, are what binds us, so read them.
On the product side, Know keeps your facts in one record, and Hue puts anything public, priced or paid in an approvals queue for a person before it goes out, which gives you the human review that this whole area keeps pointing back to. Each feature on a module page is labelled Live, Early access or Coming. None of this is a guarantee about ownership or about any legal outcome. See also The Agentic Enterprise.
Sources
- U.S. Copyright Office, "Copyright and Artificial Intelligence": Part 2, Copyrightability (January 29, 2025), and Part 3, Generative AI Training (pre-publication, May 9, 2025)
- Thaler v. Perlmutter, No. 23-5233 (D.C. Cir. March 18, 2025)
- Government of Canada (ISED), "Consultation on Copyright in the Age of Generative Artificial Intelligence: What We Heard Report" (February 2025)
- Légis Québec, Act respecting the protection of personal information in the private sector (P-39.1)
This post is general information, not legal advice.