Using AI Can Get Your Company Sued

An actual attorney explains it to me- and you.

This is my personal journal about my learning journey in AI and ML for marketing. With over twenty years of experience managing and marketing tech companies, I have deep knowledge of marketing analytics and attribution, always focused on generating measurable and meaningful business results. For the last nine years, I've advised and served marketing leaders in tech companies, enabling me to share my insights on the impacts of AI and ML, including privacy and ethics issues, through this Substack. You can learn more about me, including my policy on using AI for this blog, in the "About" section of this Substack. I welcome your suggestions for further education on AI and ML.

In This Issue:

• Brent Britton, esq. talks to us about AI and how to think about it as marketers.

• Playing with AI Generated Images

• Celebrating Humanity

• What’s Next

• A Question for you
  

Brent Britton, esq: Attorney. Engineer. IP Expert. Pianist. Badass.

Brent Britton is Fiddlehead’s attorney, but the reason you should listen to him is because he’s also (as far as he knows) the only attorney to graduate from the MIT Media Lab. He’s graciously agreed to answer some questions for me - and you- “off the clock.” There’s so much speculation from non-lawyers out there, that I am thrilled he decided to go on the record and give us all some good, professional advice.

Brent gave me no consideration for this post, it’s simply a gift he’s giving to you as part of my community. If you need a good attorney, you should hire him and the team at CoreX Legal. Fiddlehead is on the smaller end of their practice, but we have been very satisfied with their patient, wise, and practical counsel; plus I guarantee you that your attorney’s photo is nowhere as cool.

Brent, Tell Us About Yourself.

I have, uh, brown hair, blue eyes. I enjoy surfing, backgammon and men who aren't afraid to cry.

But seriously folks, I’m just this guy, you know?

Ok, Fine. I’ll be specific: What is your area of legal expertise, and where are you in your understanding of AI and the law today, especially as it pertains to marketing content (text and images)?

A million years ago, before I went to law school, I was an engineer. I studied AI as an undergrad, and also as a graduate student at the MIT Media Lab under Marvin Minsky, basically one of the gods of AI. True conversational AI has been “ten years away” for 50 years.

Alan Turing proposed a test in the 1950’s: if you are communicating with an AI and you cannot tell the difference between the AI and a human, the AI passes the test. We did not credit AI with passing the Turing test when it beat a grandmaster at chess, or when it beat humans on Jeopardy. But last November when ChatGPT went public, most people would say AI has now passed the Turing test.

The legal climate at present is in a tizzy. AI can only learn by reading and analyzing existing stuff. But if you review my stuff and produce substantially similar stuff, that constitutes a copyright infringement. So it is likely that the deployment of AI constitutes a copyright infringement of massive proportions.

Before we go on, it’s important to note that you are not offering specific legal advice to me, to my company, or to anyone else who reads this piece, right?

That’s right. To hell with you freeloaders. If you want meaningful legal advice from me, you’ll have to pay for it! And don’t go trying to offer any personal favors if you get my drift because I’m a very happily married man. You won’t catch me doing *that* again. (editor note: now you see why I like him so much. I have no idea what favors he’s talking about, but I wouldn’t be surprised if it happened.)

How similar or dissimilar is the AI climate to the time when Napster roamed free?

In the 1990s, there was no AI in common use. Napster had nothing to do with AI. AI was not a real thing yet – it was a laboratory thing.

What is the risk to brands and agencies if creators use generative AI without disclosure?

If an AI is composing your copy, no one owns it.

Ownership of words is governed by copyright. Copyright authors must be human. If your AI writes it; then you don’t own it, the AI doesn’t own it, and it is in the public domain. This means that no one owns it. Anyone can copy it and repurpose it.

 *But* if the copy produced by the AI is substantially similar to any pre-existing works that the AI had access to before it composed your work, then *you* may be liable for copyright infringement. Not the AI – AI’s are not people and cannot be sued. Just you, the user of the AI that produced the substantially similar, infringing content.

That’s Terrifying! Is generating a first draft then editing it enough to gain copyright protection? Practically speaking, how much should we be changing the copy that AI spits out, in order to protect our companies from copyright claims?

So that it is mostly unrecognizable from the copy. The result should not be substantially similar to the work that was copied.

(Editor's note: I pressed him on this several times. He didn’t budge. He’s not kidding.)

Should brands and agencies be monitoring AI usage at work? What are the key questions to consider?

Yes, and the rule is: DO NOT PUT AI-GENERATED CONTENT INTO OUR WORK PRODUCT! And, every company should have an AI policy.

How can brands protect themselves from this risk when hiring agencies, contractors, consultants, and employees?

Reps and warranties and indemnifications. You’re looking for language that sounds like this: “Writer represents and warrants that all deliverables are the original works of Writer and that Writer has the absolute right and license to deliver the same to (me). Writer will indemnify (me) against any claims arising from a breach of this warranty.”

(editor’s note: check your Master Service Agreements and Professional Service Agreements. Get addendums if you must. It’s very likely agencies are already using generative AI in your deliverables if you are not specifically asking otherwise. )

If employees, contractors, consultants, or agencies use Chat GPT-is this a potential violation of their NDA?

Everything they type into ChatGPT prompts is a breach of any NDA they’ve signed- if what they type into the prompt contains confidential information governed by the NDA. (Editor’s note: this includes spell checking documents that contain confidential information. Your internal policies and training should cover the use of AI.)

If you and a competitor are both using AI, what kind of protection and/or liability do you have if accused of copyright theft?

There is no “everyone’s doing it” defense to copyright infringement claims. If you are using a tool that has copied other people’s stuff and is producing infringing copies, you’re liable.

How should multinational companies, or those regulated outside the USA, manage their use of generative AI?

They must tune their practices for each jurisdiction.

Ok, I’m sufficiently freaked out. How can marketers and agencies use AI in an ethical and legal way?

Use it! Chat with it! Explore it. Integrate its results into your work. Just have a policy for its use, do NOT let it be the sole author of your work, and do NOT give confidential materials to the AI.

Anything else you want to add?

Ok look, my answers above may seem draconian, but the truth is we are still figuring all of this out legally. There are numerous cases pending and we’re not sure which way they will turn out. Many of the AIs available today were trained on data – software, photographs, etc. – that are protected by copyright and subject to licensing terms. If the owners of those data can see substantially similar copies in AI output, then they will be likely to keep suing. At present, they are suing the AI publishers themselves – OpenAI, etc. But they might decide to sue you, the user, just as easily.

So here is the practical advice:

1. Be careful. Use AI to produce stuff, but go over it with a critical eye. Try to make sure the AI output you’re using isn’t too similar to any input that may have been used to generate it. Not easy; I get it. But if you see a Getty Images watermark on the AI output, that’s a pretty good tell.

2. Don’t forget that only humans can be authors and inventors, so you do not own the AI output. No one does. It’s all in the public domain.

3. Remember that everything you type into the AI prompt is no longer confidential and may be fed into the AI for future use. For goodness sakes, read the terms of service that comes along with your AI of choice and ensure you understand what it says about ownership and confidentiality and especially indemnification because it probably says that if the AI publisher gets sued because of output you generated using AI then you have to defend that case. Fun times. (Editor’s note: For him, not for you.)

4. If you are using AI to generate software, be very afraid. One of the earliest lawsuits was filed over massive software copying; it is still pending.

5. Finally, remember that the AI is full of sh*t and will lie to you all day. It will make up facts and then insist that it is telling the truth. It will invent fake bibliographical citations that look very real but will hold you up to ridicule if you publish them without fact-checking. This dynamic is lovingly called “hallucinating” in the parlance, as though the AI were just temporarily tripping or something. Ha ha… so cute.

TL; DR: AI is a great tool that will make your work better, stronger, faster, and more efficient. But you must review and scrutinize every single thing it produces before putting that output into production because it could be wrong, or flat-out made up. It could be an infringing copy of someone else’s work. It could subject you to liability. Let’s be careful out there.

Editor’s Note: want to hear more from Brent? He did a webinar this week. Check it out.

Celebrating Humanity:

As you know, I have been undergoing breast cancer treatment for the past year. As I write this, I am celebrating the one-year anniversary of the first phase of my treatment, which was surgery to remove the tumors in one breast, as well as several lymph nodes. Thankfully, the lymph nodes turned out to be negative. I’m considered NED (No Evidence of Disease), and if the statistics apply to me, I will be around for a normal lifespan.

Now, if you had met me prior to my cancer diagnosis, you would know I had "plenty to spare" in that department, and frankly, the idea of a mastectomy was appealing: who wouldn’t want to skip wearing a bra again? And if it ensured that I was done with cancer once and for all, it seemed like a fair trade. However, Dr. Susan Love (via her book) did the first research that uncovered this: mastectomies (in most cases, but especially in my case) have no better outcomes than a lumpectomy.

Prior to Dr. Love’s work, women would go in for a lumpectomy and come out with a double mastectomy without consent. Her book is the bible for most of us with breast cancer. Her untimely death on July 2 from recurrent leukemia was a loss to all of us in the breast cancer community. She was a pioneer in surgery, same-sex marriage, and just being a woman in a very male-dominated environment. She went on to create a foundation and an army (literally, we're all part of Dr. Love's army of patients who volunteer for clinical research).

You owe it to yourself, to me, and to anyone you care about who’s had breast cancer, to read her obituary. I’ve gifted a link for you from the Washington Post.

Playing With Midjourney Bing and Substack Images

A simple word, summer. And I’m so grateful this year to be able to enjoy it.

summer

What’s Next

• I’m off to MAICON: Marketing AI Conference Next week. I’m sure I’ll be fired up with lots of ideas upon my return. Keep an eye on my Notes here, as I will be posting interesting tidbits. I will also post business things on LinkedIn. I’m on Threads and Bluesky, but neither of them feel like a good use of time right now.

• Please continue to share with your friends. It keeps me inspired and excited to find out new things.

• Season 4 of Stayin’ Alive in Tech is going to be all about AI. Who do you want to hear from?