In our years of advising marketing agencies on legal risk, we’ve seen lots of economic, social and technological developments that continually change the business game for them. None so far has moved so quickly to force adjustments to the ways agencies think, operate and serve as has AI.

This, regrettably, doesn’t mean that the law evolves at equal pace.

Fortunately, though, the last month has brought some developments around how copyright law might evolve in reaction to the realities of AI. We’ve also got some signals about how lawmakers might approach regulating issues like data privacy and protection issues related to AI use.

What are the updates, and what does it mean for your agency and your clients now?

1. What’s changed around AI and the use of copyrighted works?

We have been waiting quite a while to get additional clarity around the legal protection of copyrighted works that are fed into AI platforms. While the U.S. Copyright Office advanced a position that seemed highly protective of copyrighted owners in early 2025, there were several pending court cases around this issue that “broke free” earlier this Summer.

First, let’s back up a second and review the two primary questions marketers have on their minds when it comes to generative AI and use of copyrighted work as inputs, or the legal risks of using AI output:

1) Are you permitted to do this at all, particularly for AI model training purposes, and

2)  Can you be liable for copyright infringement if you create and use output from generative AI that is substantially similar to someone else’s copyrighted work?

For now the answers appear to be Yes, and also……Yes.

Courts in two recent federal case decisions (Bartz v. Anthropic and Kadrey v. Meta) both allowed copyright-protected materials to be used to train AI models. The TLDR version of key copyright issues addressed in these opinions centered around whether the use of someone’s copyrighted work is “transformative” (meaning the new use of the work (like data training) is so different from its original purpose (such as to entertain) so as to be considered a new work for copyright purposes). The courts in both these cases said yes – it’s transformative, and therefore not infringing, to use someone’s copyrighted works to train an AI model.

There’s a lot of fine print around the courts’ decision-making in these cases, and a lot of factors the courts deliberately did not consider, so let’s focus on what the marketers in your agency need to be thinking about now.

• What does this mean for agencies?

There are a few things to keep in mind in light of these recent court decisions as you use generative AI at your agency to create concepts, outlines or work product for your clients:

• There will be a lot more copyright-protected works input into AI, many of which will show up (now, potentially lawfully) in the output/results that creatives at the agency might incorporate into client deliverables. This raises your risk of getting output that infringes on the copyrights of other parties.
• There will be additional due diligence required to clear final work and assets for potential copyright infringements – and in some cases clearance will not be possible.
• The terms of the AI platforms your teams are relying upon are not likely to reduce your liability for copyright infringement. You’ll need a mechanism in your client contracts to limit your liability for missteps, and some clear policies around asset approvals, to reduce agency legal risk.

If you’re an agency training your own AI models, using properly licensed or, ideally, agency-created or other owned content to train the models is still the safest practice.

2. What hasn’t changed (yet) about IP infringement liability associated with using AI-generated content?

We’ve sort of spoiled the lead here. But let’s be specific about what hasn’t changed so far around copyright, trademarks, and generative AI.

Despite the fact that AI platform owners may be able to (lawfully) use copyrighted works to train their models, there is still no safe legal pathway to using output from AI that accidently infringes the copyrights, or trademarks, of another party.

One key pending court decision on this issue is the Disney v. Midjourney case. Disney is not happy that Midjourney, an AI platform, knowingly allows users to reproduce well-known characters owned by Disney that have been input into its training data. Midjourney’s position is that it’s not their fault if users take output and use it in an infringing way. The resulting holding could impact how court’s look at “accidental” or unintentional IP infringement that arises from AI output.

• What does this mean for agencies?

Your agency’s risks around using AI output in a way that potentially creates an infringement of someone’s IP remain intact – even as AI platforms might now be inputting more copyrighted works into their data.

So what can you do to minimize those risks?

• Keep your protocols intact for vetting IP use – this is no time for panic, or for false security. Performing trademark searches, copyright clearances, and ad copy reviews needs to still be a part of a marketer’s protocol before releasing work into the world. You should already have these protocols in place regardless of your level of AI use.
• Have clear conversations with your clients about the responsibilities for IP due diligence. One conversation should be about who will assume responsibility for the due diligence around IP use, and the approvals of all assets before publication. The second discussion needs to be around the assumption of liability if an IP issue around AI use goes sideways – and this conversation should be reflected in your contract terms.
• Be highly familiar with the legal terms of the specific platforms your agency uses, including the level of subscription (some tiers of paid access to the platforms may offer enhanced assurances or protections around IP issues – rely on them at your own risk).

3. Are there any upcoming laws or regulations agencies should be aware of?

Many AI-related bills have been introduced into both state and federal legislatures over the last year – far too many to cover in one post. That said, there are two federal bills that have been introduced recently that could substantially impact agencies’ use of AI.

The first of these potential laws is the AI Accountability and Personal Data Protection Act. This law would allow individuals to sue AI developers for using their copyrighted content or personal data without their consent. It would also require disclosures of how AI models handle third-party data.

The second of these potential laws is the Generative AI Copyright Disclosure Act, which would require companies to submit lists of copyrighted works used to train gen-AI models to the US Copyright Office before the model is publicly released. This contemplates a pretty cumbersome process for AI platform owners.  Not everyone is going to have (or take) the time understanding, or ability to get clearance from the Copyright office.

• What does this mean for marketers?

The overriding theme on the regulation front, so far, is that we’re feeling a very familiar push-pull on the balance of federal vs. state lawmaking efforts to regulate AI. This creates the same kind of patchwork environment agencies have to manage for years now when it comes to data privacy issues – no overriding federal rules, and a lot of state laws to manage.

Despite that, the most likely impact on the lawmaking front seems to be related to data privacy compliance when using AI platforms to manage, interpret or manipulate data for marketing purposes. We’d actually welcome some federal standards here on behalf of our agency clients – one set of laws is much less confusing to comply with than dozens of state regulations.

Expect the regulations and case law decisions around AI use to evolve regularly, but more slowly than the technology is affecting your agency’s business in other areas like operations, workflow, and creative output.

We’ll keep you posted regularly as the legal landscape evolves, but if you have questions meanwhile about what the recent updates mean for your agency, we’re an email or phone call away.