OpenAI Misfires. Will it Follow My Rules for Litigation? Probably Not. But You Can!
OpenAI Creates Bad Law and Jesse's Rules for Litigation
Welcome to the 11th edition of my newsletter. Last week, I was called for jury duty! I didn't have to serve this time, but between that and representing my client in a limited liability company membership dispute that has devolved into litigation here in New York County (Manhattan), I’ve spent more time at the New York State Supreme Court, New York County courthouse over the past few weeks than I did in the prior 8 years of being in-house counsel. Everything old is new again, I guess.
Being in the courthouse that much made me reflect on litigation, aka “dispute resolution through the courts,” especially how small-ish businesses should think about litigation risks. Below are my rules for litigation gained from clerking for a judge and from 20 years of litigating cases big and small at law firms big and small, on both sides of the “v.”
But first, my take on the big litigation du jour: Scarlett Johansson v. Open AI.
Can a Robot Shoot Itself in the Foot? OpenAI Proves Its Possible.
OpenAI and Scarlett Johansson! This is one of those rare sort-of-wonky legal stories that actually broke through to the mainstream consciousness, and everyone has a take. But here is mine: I’ve written before about some of the challenges presented by trying to apply traditional legal principles to artificial intelligence. They don’t really fit.
As I’ve written about in the past, anyone can, if they want, get one of these chatbots to say something defamatory. The key issue in the pending defamation case against OpenAI is whether that private chat constitutes a publication by OpenAI. I don’t know how the Court is going to think about that—clearly, the chat was supposed to be read by someone, and the AI was not talking to itself, but is OpenAI the defamer, or is the user?
In Scarlett Johansson's case, OpenAI created a voice for its chatbot, apparently using a voice actress that sounded eerily like Scarlett Johansson’s voice from the movie “Her” after Scarlett herself declined to license her voice. For a “misappropriation” or “publicity” claim, it's a really bad set of facts for OpenAI, and I expect them to want to settle just to put this wholly strange episode behind them. Typically, I would say that licensing the voice as part of a settlement is possible, but this has been a debacle, and that is likely not even an option. But really, what's wrong with that eerie Google Maps voice, or Alexa for that matter? Maybe ChatGPT can just sound like that.
OpenAI is going to settle the case, and the terms of the settlement will be confidential, including the amount. But that amount is going to be a lot. And that is not because Ms. Johansson would win more money at trial, but because of the precedent she could set, which would be really, really bad for AI companies. And cases like this are going to keep happening, over and over.
Keep in mind that these are not copyright cases because “visual styles” and “voices” can’t be copyrighted. These are “publicity cases” AKA “unauthorized use of identity” cases. OpenAI’s mistake here was not trusting their own technology.
Why? If you ask ChatGPT, or any other chatbot, to write a song about your dog in the style of Taylor Swift, you’re going to get that. Does Taylor Swift have anything to do with ChatGPT? No. And the fact that you can’t do it for video or voice is just a technical limitation. Make me a video clip about my dog in the style of David Lynch? It’s coming. Read this poem about my dog in the voice of Arnold Schwarzenegger? It’s coming too. Did anyone at OpenAI code its LLM to understand how to mimic these celebrities? Probably not. Does it matter? Yes, it might.
Had OpenAI developed their own sexy-sounding female voice by training their AI on all the open-source videos and recordings with sexy female voices, and then hired some sound engineers to tweak it, and it happened to sound like Scarlett Johansson, then we would have an interesting case. Sort of like having the AI write a song in the style of Taylor Swift or a movie in the style of David Lynch or a voice that sounds like Arnold Schwarzenegger. Does that violate those celebrities' rights? We don’t know. That case is coming.
But what happened here is that they asked for permission, were refused, and then hired a voice actress to sound like Scarlett Johansson. Oops. These are bad facts. One could say that OpenAI should not settle because the case is so extreme that every other case—like the apocryphal Taylor Swift case—will always be distinguishable. But I’m not so sure. If you program a malicious bot, claiming it went rogue is unlikely to get you off the hook. Yes, it's even weirder when you dress like a bot and act maliciously, which is sort of what OpenAI did here when they just hired a voice actor to mimic Scarlett Johansson. And so what most people, including legal people (“lawyers”), are going to take away from OpenAI’s likely loss is that if you build AI and it mimics someone famous, you’ve violated their publicity rights even if you’re not so intentional about it. If you happen to be intentional, like OpenAI apparently was, you just look stupid.
And this is where the law around AI is headed. Unlike social media companies, which can rarely be held liable for user-generated content under Section 230 of the Digital Millennium Copyright Act, AI companies don’t enjoy the same protection because the content is not user-generated. Your robot created that – it's the whole point. And no one is going to give AI a legal shield anytime soon. And so OpenAI has just shot itself in the foot. Unless, of course, they pay, settle, and wait to get sued again where the facts are better. Which is what I expect to happen.
Remember how I said “that case is coming?” That is key. If OpenAI was my client (they are not), I would have advised them to handle this very differently. Assuming that getting Scarlett Johansson’s voice was some type of imperative from the product team, advising them not to create an approximation of it using a voice actress was never an option and no good lawyer should just think that a simple “no” is going to solve anything. Further, let's assume someone actually called Scarlett Johansson to ask if they could license her voice and she said “no” (or was non-committal). And further assume that person was the CEO of the company. Now the product/voice is out there, and Scarlett Johansson is threatening to sue. This may seem strange, but I’ve dealt with similar bad facts.
OpenAI’s lawyers did something right: they took down the voice. Damages in these publicity cases are the greater of actual damages or $750 per use in California, and even before looking at how the $750 per use is calculated (is this one use, or a million uses based on the number of sessions?), just knowing that there are statutory damages is enough to make me say: pull it. But OpenAI’s lawyers then made a huge mistake because they let someone else botch the messaging by claiming, essentially, that this was all a misunderstanding. The actual statement was “We are sorry to Ms. Johansson that we didn’t communicate better.”
That was the wrong approach. OpenAI should have said yes, we created a voice that sounded exactly like Scarlett Johansson. Yes, we have the right to. Why? Because this is not a publicity case at all and this is the way the creative process works - Scarlett Johansson inspired us the same way someone inspired her to create that voice. Don’t even accept the premise that we did anything wrong. Why did we call? We were thanking her for the inspiration.
The end result would have been the same - Scarlett Johansson is getting a fat check. But at least OpenAI would not look like a bunch of amateurs and the defense to the next lawsuit, the real one where your AI actually impersonated someone all by itself which is, after all, what the AI is designed to do, would be teed up. This is a long game and a smart lawyer is always a step ahead. Dry powder, and no bullet hole in your foot. You’re welcome. (Oh wait, did I misappropriate?)
At that, here are my rules for litigation.
Litigation: The Worst Way to Resolve a Dispute, Except for all the Others
Dispute resolution through the courts (“litigation”) is actually what I trained to do, and up until about 2013, I was pretty much exclusively a litigator. Having worked as a law clerk to a federal judge, litigated as part of, and against, firms big and small, in matters big (billions of dollars at stake) and small (thousands of dollars and lots of ego at stake). I’ve also hired and managed outside litigation counsel from large firms and small firms, for matters large and small. I have a lot of perspectives and experience. And, of course, any smart business is always thinking about their litigation risks, and I’m always advising my clients (a smart business, everyone of them).
These days, litigation is about 40% of what I do. On most litigations, I work as Of Counsel to Harwood Law PLLC, a commercial and corporate and employment litigation firm run by my friend Tony Harwood, one of the finest litigators in New York City. I guess that is rule number 1: never litigate alone.
But really, I have a few other rules for litigation that I’ve developed over the years of being on the plaintiff side and the defendant side. Call these “Jesse’s Rules of Litigation.”
Avoid Litigation, But Don’t Fear It: Lawyers, and especially litigators, are great at bluster. Drafting a demand letter designed to scare the crap out of someone is a pretty common lawyer trick (note to anyone who got one from me and might be reading this: I mean business!). If you get one, stick to your guns and your position. Never ignore the demand and hope the issue goes away. If there is something to be done before the litigation commences, by all means do it. If you want to tell the lawyer to take a long walk off a short pier (or as they say in Yiddish “gay kaken ofn yahm” (look it up yourself)) and then wait to see if they make good on their threat, that is a viable option (some lawyers – not me – like to “fish” for easy settlements). Don’t capitulate because someone threatens to sue you. On the flip side, never try to use litigation to bully someone into doing something unless you have a real case. That never works.
Indignation and Righteousness Are Not Strategies: If you’re a defendant, it's natural to be super angry you’re sued. But indignation should not be your legal strategy. If you’re a plaintiff, it's natural to think you’re a righteous litigator suing to make the world right again. I often get on calls with opposing counsel where words like “extortion,” “liar,” “principle,” “fairness” are bandied about. Talk is cheap and indignation and righteousness do not win cases. Preparation, consistency, and solid legal strategy are the key to victory.
Resolve a Dispute, Don’t Perpetuate It: When I’m involved in litigation, everything I do is geared toward getting the case resolved efficiently. Once you're sued or you make the decision to sue someone, stick to that claim and don’t try to widen it to encompass more. It's fine to bring a counterclaim or to bring related claims, but don’t move too far off the initial dispute. So if someone sued you for non-payment, it's OK to counter-claim for non-performance (the reason why you did not pay is because they did not perform, after all). But if you’re sued for non-payment and you counter-claim for non-performance and throw in some type of emotional distress claim because the allegedly false allegation that you did not pay was so hard on you, and you want to sue them for defamation because they left a nasty review saying “he does not pay his bills,” resist the urge. Remember, your goal is just not to pay for something that you don’t feel you have to, not to recover affirmative damages. The temptation is to think that you can bully the other side into dropping their claims or settling a claim (which never works, see above), but in reality all you’ve done is (1) bring weak defenses or claims that diminish your strong ones; (2) hardened feelings; and (3) cost everyone more money. Another common approach is to try to go after the lawyer with sanction threats, alleged ethical violations, disqualification, etc. That is rarely productive and is typically counterproductive since you lose credibility when you don’t follow through on the threat and give the other side momentum when they prevail. Those cases are losers 90% of the time and courts hate dealing with disputes among lawyers.
Decide Whether You Need to Outthink or Outwork the Adversary: This is more for the lawyers than the clients, but it should impact how you choose a lawyer to represent you. If you’re a small company sued by a big one, and they bring out the “big guns,” you need to outthink them, because you will never outwork them. They have the resources to just keep litigating, and they will (as long as the bill is paid). But the opposite might not be true: If you’re sued by a little firm (which is common for employment discrimination) you may be tempted to hire a firm that will bury the other side in motion papers, discovery demands, whatever. That does not work because little firms get lots of deference and courts hate bullies. Of course, there are some instances where you always need to outwork the other side. That is generally the case when the other side just sort of sued thinking they could extract a settlement in a commercial dispute. In those cases, it's best to just outwork the other side until they realize that getting anything out of you is going to be a very expensive proposition.
Know Your Insurance: Insurance coverage and litigation always do a super complicated, but very important, dance. Insurance won’t cover most contract disputes, but it will cover you if someone is claiming your product hurt them and for “advertising injury” which is pretty broad and could cover lots of IP-related claims (OpenAI’s lawyers are getting a crash course in that at the moment). If you’re a fast-growing startup hiring (and firing) lots of people, an Employment Practices Liability (“EPLI”) policy is probably a good idea because, in my experience, the most common lawsuits that startups face are employment-related. Assuming you have applicable insurance, deciding whether to report the claim is an important decision. Insurance companies will generally pay “first dollar defense.” Sometimes the first dollar defense is subject to the deductible, sometimes not. It's important: if it is not, you’re going to get a bill from the insurance company for your defense even if you win completely. So if defense costs are subject to the deductible and you can get rid of the case for under the deductible, it might not be a good idea to get the insurer involved, since it's going to influence premiums going forward. If it's something that is definitely going to settle above the deductible they certainly report it and get counsel assigned by the insurer and have them pay the bill - you just basically got a free lawyer. The decision on whether to use “panel counsel” or your own is another decision. For something like a slip and fall in a place where your typical counsel is not licensed, the panel counsel is going to be fine. But if you can, it's always better to work with a lawyer you know and trust. You should ask your lawyer to either reduce their rates to the insurance company rates or agree on a (reasonable) amount of a supplemental rate that you’re going to pay. If your lawyer tells you they won’t work for the insurance company rates even though they would be the best counsel for you and you want them, start working with another lawyer. Really. The rates are not awful, and the relationship should be more important than the money. Just a note: Insurance really only comes into play when you’re being sued. The flip side of insurance is the relatively new field of “litigation funding” where you can find someone to essentially invest in your lawsuit. It's an interesting field, and I’ll write about it in a future edition of Fractionally Yours.
Understand the Pacing of Litigation: Litigation moves slowly. That is good for defendants and bad for plaintiffs. Sometimes the party sued, the defendant, really, really wants vindication and wants to move things along while the plaintiff – who typically wins the first few rounds and scores points just by suing – just sort of wants to slow things down and savor the win. That is exactly wrong. Plaintiffs should want speed. Keep pushing, always be ready. Defendants want things to go as slow as possible. And always make sure that your litigation is not interfering in the operation of your business. If it is, whether you're a plaintiff or a defendant, you’ve already lost.
Finally, None of Those Rules Apply When Litigating Against the Government: Litigation against the government, either as a plaintiff or a defendant, is a totally different animal. The government has unlimited money and does not care about winning or losing, just doing “justice.” But on the flip side, you can see them coming a mile away and have time to adapt and avoid litigation. Unless it's existential, I always advise my clients to defer to the government (also known as “doing the right thing”).
Those are my rules. If you have some suggestions based on your own experiences, please let me know!
Keep thinking, keep building,
Jesse
Hi, and welcome to my newsletter! I’m Jesse Strauss, Your Fractional General Counsel. I’m a lawyer with a private practice based in New York City, helping clients in the United States and globally with their US legal needs. My expertise spans various areas, including raising funding rounds, employment issues, negotiating master service agreements, intellectual property, compliance, legal process management, and dispute resolution. My focus is on founding and nurturing great companies from seed to exit. Discover more at www.yourfractionalgc.com and book a complimentary 30-minute consultation at Contact Your Fractional GC.