Who Owns What AI Creates? What Business Owners Need to Know
At our October 7 AI for Business workshop at Roseville Venture Lab, Anders Bostrom asked how people were using AI. Hands went up for writing emails, researching, creating marketing content, and generating code.
This was a room already putting AI to work. The harder question was what happened after it produced something useful: Could you protect it? Could you use it commercially? And what had you shared with the AI company along the way?
Those questions drove “Who Owns What AI Creates?”, a discussion with Bostrom, a Senior Associate at Feldesman LLP whose practice includes corporate and healthcare transactions, commercial agreements, and regulatory compliance.
Bostrom described himself as an AI optimist. His goal was to help business owners understand the obligations that come with tools they are already using.
Here are the lessons that stood out.
1. Permission to use an AI output does not guarantee you can protect it
One of the most useful distinctions of the evening was between what an AI company’s contract promises and what intellectual property law protects.
A platform may give you rights to use its output or agree that it will not claim ownership. That does not automatically mean you can prevent someone else from using similar material.
For a business owner, these are three separate questions:
Can I use this? Check the tool’s terms and the plan you purchased.
Can I stop someone else from copying it? That depends on the legal protection available.
Could my use infringe someone else’s rights? The platform’s permission does not settle that question.
Consider a founder who generates an illustration for a marketing campaign. Permission to publish the image does not necessarily provide exclusive rights to the image.
That distinction matters more when the output becomes a valuable business asset.
2. Your human contribution matters
Bostrom emphasized that U.S. copyright protection centers on human authorship.
The U.S. Copyright Office’s guidance explains that using AI as an assistive tool does not disqualify a work from copyright protection. Human-written text, creative arrangements, and sufficiently creative modifications may qualify. Purely AI-generated material does not, and prompts alone generally do not establish authorship of the resulting output.
The scope of protection depends on the actual human contribution. Changing a few words is not a guaranteed way to make an entire AI-generated article yours.
For founders creating valuable content, Bostrom recommended preserving drafts and documenting their creative work. Keep your original writing, sketches, revisions, and decisions about how the material is organized.
Think of this as keeping the working drawings alongside the finished product. They help show what you created.
3. Copyright and trademarks protect different things
An audience question about AI-generated logos highlighted another common source of confusion.
Copyright concerns original creative expression. Trademark protection concerns the identifiers customers associate with a business, such as its name or logo.
An AI-generated logo may still function as a trademark if it meets the applicable requirements. Its AI origin does not eliminate the need to check whether it conflicts with an existing brand.
Bostrom’s recommendation was to clear names and logos before investing in them. The USPTO similarly recommends a comprehensive search for confusingly similar marks, including uses beyond its federal registration database.
Imagine generating a name, buying the domain, printing signs, and then discovering that a similar business already uses a confusingly similar brand. The inexpensive logo can become an expensive rebranding project.
AI can help brainstorm. Trademark clearance needs its own process.
4. Pay attention to what goes into the tool
Some of the most immediate business risks involve the information you upload.
Bostrom discussed source code, customer information, business strategies, and client documents. These can carry confidentiality obligations or represent information your company needs to keep secret.
Uploading a document to a cloud AI service means involving another provider in processing that information. Whether that is permitted depends on your agreements, the provider’s terms, and the safeguards in place.
Before uploading sensitive material, ask:
Does our client agreement or NDA allow this use?
Can the provider use the information for model training?
How long is it retained, and who can access it?
Is this a company-controlled account approved for this work?
Bostrom encouraged businesses to evaluate business and enterprise accounts for stronger contractual protections and administrative controls. But paying for an account is not, by itself, a confidentiality strategy. Review the specific plan and settings.
The same review should extend to AI connected to email, customer databases, and document storage. A connector can expose information without anyone manually pasting it into a chat.
5. Hiring a contractor does not make the AI questions disappear
A founder may carefully manage their own AI use while overlooking the agency designing their website or the contractor writing their software.
Bostrom encouraged businesses to address AI use directly in those relationships.
Ask which tools the contractor uses, what company information enters those tools, and which parts of the deliverable are AI-generated. Discuss ownership, licenses, source files, and disclosure requirements before work begins.
His slides also cautioned against assuming that calling contractor work “work made for hire” resolves ownership. Appropriate written IP assignments and other contract provisions deserve attention.
A contractor cannot transfer copyright that does not exist. Paying an invoice is not a substitute for understanding what rights come with the deliverable.
6. Review outputs before they become your business’s problem
Bostrom recommended human review before AI-generated material is published or shipped.
For marketing, that means checking for recognizable brands, protected characters, misleading endorsements, or other material that creates risk. For software, it includes checking whether suggested code carries license obligations.
Reverse-image searches, plagiarism checks, and code-scanning tools can help identify concerns. They are checks, not guarantees.
He also advised reviewing any IP indemnity offered by an AI provider. In plain English, that is a contractual promise to cover certain intellectual property claims, subject to conditions and exclusions.
Understand what triggers that protection, what is excluded, and whether required safeguards must remain enabled. A reassuring headline on a pricing page is not the whole agreement.
7. Treat AI meeting notes as a business decision
The discussion continued into AI meeting assistants, recordings, and automated notes.
Bostrom recommended disclosing recording or transcription and obtaining consent before starting. California generally requires all parties’ consent to record confidential communications, although the application depends on the circumstances.
He also raised questions beyond consent: Where does the transcript go? Who can access it? Should an outside note-taking service attend a confidential client or board meeting?
An AI summary can miss sarcasm, qualifications, or the difference between an idea being explored and a decision being approved. For formal governance, it should not replace the process of reviewing and approving minutes.
Before enabling a meeting assistant by default, decide which meetings it belongs in and how its records will be handled.
Pick one improvement this week
Bostrom returned to a practical challenge throughout the workshop: identify one step that would better prepare your business.
His closing recommendations offered five places to start:
Review your AI accounts. Evaluate approved business tools and their confidentiality, training, and retention terms.
Create a one-page AI policy. Define approved tools, prohibited inputs, required review, and when disclosure is necessary.
Update contractor agreements. Address AI use, confidentiality, ownership, and delivery requirements.
Document human contributions. Preserve evidence of the creative work behind assets you want to protect.
Read your provider’s indemnity conditions. Understand what help is available if an IP claim arises.
My suggestion is to start with an inventory of the tools your business actually uses, including the ones built into other software. Then choose the biggest gap you can realistically fix this week.
AI is already helping small businesses produce work they previously lacked the time or budget to create. Understanding what you can use, what you can protect, and what you need to keep confidential helps you put that capability to work with fewer surprises.
Thank you to
Anders Bostrom for sharing his time and answering the community’s questions. Visit Startup Folsom for upcoming workshops and opportunities to learn alongside other founders and business owners.
This article summarizes an educational workshop and is not legal advice. AI law and vendor terms continue to change; consult a qualified attorney about your specific circumstances.
Sources and workshop resources
About Anders Bostrom
Anders L. Bostrom is a Senior Associate with Feldesman LLP. His practice focuses on advising organizations on complex transactions, regulatory compliance, corporate governance, operational matters, and business agreements.
Anders works with healthcare providers, physician groups, management services organizations, health systems, healthcare technology companies, private equity sponsors, and investors. He has been recognized as a Northern California Rising Star by Super Lawyers from 2023 through 2026 and was named to The Best Lawyers in America: Ones to Watch 2026 for Corporate Governance and Compliance Law and Corporate Law.
