ScarJo, ChatGPT and Me

Working in advertising and entertainment, my friends sometimes ask me, "Who is the most famous person you've met from work?" My immediate answer is always "Scarlett Johansson!" And then I have to backtrack, "Well, I didn't actually meet her, but I had lots of conversations with her lawyers!" And then I have to backtrack a bit further, "Well, I didn't actually have conversations with her lawyers, but we exchanged a lot of really lengthy emails!" My point here is that no matter how excited you may get, it's important to get it right (eventually) when talking about ScarJo.

And that brings us to the past few days. According to Scarlett Johansson, OpenAI asked ScarJo (twice) to be the voice of its voice assistant (named Sky) for ChatGPT-4o. She declined. Many thought the voice used still sounded a lot like "Her," while OpenAI claimed that any similarities were incidental. OpenAI has since suspended the release of Sky. A couple of lessons here:

• When asking permission to use anything (a trademark, a voice, a neighbor's hose, etc.), be prepared to abandon the project if your target says "no thx." If you instead go ahead and do it despite that rejection, not only have you already put the target on notice, but it makes it harder to argue you didn't actually need permission, or that target wasn’t at least part of your plan;

• If you want to maintain an argument that there is no link between your end-product and the target, don't post on social media referencing that target. Here, CEO Sam Altman simply posted the word “her” on X. Did you know that ScarJo voiced the AI system in the 2013 film titled "Her"??;

• When you're already being accused of using folks' IP without permission in many high-profile actions (see, Sarah Silverman, Authors Guild, NYT, etc.), you may want to tread lightly around A-List celebrities - particularly A-List celebrities who have demonstrated trepidation about deep-fakes, and a proclivity to speak out against BIG corporations (see Walt Disney); and

• If you have a really cool product (which ChatGPT-4o seems to be!) don't do ANYTHING to take press attention away from how cool your product is. You may end up with important commentators (like the NYT and me!) focusing on all the wrong things...

Live Online Shopping: Dead or Alive?

If Live Online Shopping takes off in the US, how do we apply old rules to this new frontier?

Consumer brands acknowledge that success today is tied to having a robust online and mobile presence. During the pandemic, e-commerce took off, and hasn’t dissipated since, with online sales expected to grow to over $1.1 trillion in the US in 2023, even as brick-and-mortar retail re-emerges.  

But if China offers any prescient view of our retail future, the next frontier in the US may be in live online shopping. What is it? It’s TV home shopping on steroids.  

Live shopping events target a younger demographic’s perceived need for immediate gratification, desire to be constantly entertained, and fear of missing out. Platforms are developing and exploring interfaces for interactive events where either anyone with a phone can livestream and sell name-brand products from their home, or the platforms (or brands) hire known influencers to peddle products on more heavily-produced live shows. 

There are conflicting reports about the viability of online live shopping, but estimates have it accounting for as much as 25% of China’s retail market, or $647 billion this year. Whereas Chinese online live shopping is dominated by a few core channels (e.g., Alibaba’s Taobao Live, Douyin and Kwai), the US market is diversified and developing - leaving consumers with no single place to find their desired shopping show. Thus, right now, the livestream shopping market in the US is less than 5% of China’s market.  

Nonetheless, startups and some established platforms and brands are getting into the business - betting that consumers not only want to make a deal, but want to have a fun experience while doing it. Along with Poshmark’s live Posh Shows, the big brands of Amazon, Walmart, YouTube, Ebay (and, at a slower pace, TikTok) are all exploring and investing in this shopping outlet. This, despite Meta’s decisions in October 2022 and March 2023 to curb live shopping on Facebook and Instagram, respectively.  

It’s certainly possible that the US appetite for live online shopping won’t take off because of inherent American shopping habits and patterns. However, a good scout is always prepared. If it’s live, that means less control and review by brands and platforms.  This makes some marketers (and their lawyers) uncomfortable. Here are some legal considerations in this space:

• False Advertising?

To the audience, the advantage of a live broadcast is the sense that the content is unscripted and off-the-cuff.  But there’s little to stop talent on a live feed from getting things wrong about the product being sold (not to mention, saying horrible things in general!). “This soup was 100% made in the USA,”; “This purse has never been offered at this price before,” “This shirt is more durable than the competitor’s,”; “These shoes are as good-as-new after walking for 10 years and 100k miles!” If consumers rely on these false or misleading claims, the competitors, consumers, the Federal Trade Commission (FTC) and other regulators may have grounds to object. It is extremely difficult to check whether claims are substantiated when the on-camera talent doesn’t stick to a pre-arranged script.

The issues get even more complicated in situations where the brand doesn’t produce the show - and instead, a fan produces her own independent show, making unfounded claims as she sells her favorite brand’s products. Are brands going to be OK with an expansion of these types of shows, putting their intellectual property and reputation in jeopardy?  These scenarios could put brands at odds with the platforms who may be forced to limit authorization to only approved live channels/shows. And brands could find themselves writing a ton of cease and desist letters to individual, unaffiliated sellers. 

• Endorsement Issues?

Similarly, the live element makes it difficult to stay on top of endorsement issues. Per the FTC’s Endorsement Guides, marketers and influencers must clearly and conspicuously disclose their material connection to the advertiser if it’s not otherwise obvious from the context. If the well-known brand CEO is live selling brand products, a disclosure isn’t needed because the material connection of the CEO to the brand is obvious. When brands engage an influencer to live-sell their products, clearly these Guides apply, and the necessary disclosures are likely required. 

The livestreaming platforms may need to explore expanding their array of overlay and voiceover options to make it easier to disclose the connection in a live setting. And the brands may need to live-monitor their shows in back-end control rooms, to watch for both the false advertising and endorsement issues. Once the disclosure is missed (or false statement is uttered) in the live feed when the purchases are made in real-time, it is exceedingly difficult to rectify the failure.

• Counterfeit goods?

Many consumer product and retail brands have entire divisions dedicated to stopping the sale of counterfeit goods. If live online shopping takes off, it may put more stress on these divisions to catch the counterfeit items before the livestream is over - and the technology might not be ready for that. 

• Who is behind the sale?

It may not be obvious from each live seller platform how the money flows. The platform almost always takes a percentage, but there may be setups where one of, or both, the primary seller and brand share in the proceeds. It’s certainly possible a consumer’s purchasing decision will be affected by knowing to whom the money is going. Further, if any of the parties does not get a % of sales, it might factor into deciding how responsible that party should be for problems with the sale.  See below on that!

• Contractual enforceability?

The knee-jerk counterargument here is to provide protections via contract.  Brands can negotiate deals with influencers, and platforms can provide online terms and conditions to consumers and sellers, each spelling out responsibilities and disclaiming liability where appropriate. Unfortunately, contracts with influencers only work when the brands actually engage influencers (which will not be in every situation in this space); contracts likely cannot limit a brand’s inherent obligation to monitor and enforce its policies; contracts don’t stop regulators from getting involved; and online terms and conditions may have limited enforceability, given they are often ignored.

• So, who is responsible??

For all of these issues, as between the platform, the brand, and the influencers, who should be responsible when things go wrong? At least in the endorsement realm, the FTC has been fairly clear that brands, agencies and talent might all be held liable. The FTC has also pushed platforms to make it simpler for brands and influencers to make disclosures.  

However, if the brand is completely out of the flow of money and knowledge when its products are sold live online, there’s a strong argument that the brand should not be held responsible. It may not be reasonable to expect brands to monitor every secondary sale of their products, even via live sales. That being said, as noted above, the brands may still have to contend with reputational risks when unauthorized individuals (and platforms) promote their products in a manner the brands would not otherwise have approved.

The scope of live online shopping is what makes the responsibility question so difficult. When anyone can take an item from their closet and put on a live show and sell, it is not going to be practical for regulators and others to go after all the sellers. Thus, at a minimum, it may mean tighter scrutiny on platforms and brands to follow reasonable procedures, and post reasonable guidelines for these sellers to help them avoid these pitfalls.

___________________

As with most advertising legal issues, the issues aren’t novel, but the landscape for applying them is. And a “live” landscape moves as fast as possible, making it difficult for everyone (retailers, regulators, lawyers) to keep up. Stay tuned!

For legal questions about the live online shopping industry (or any advertising/media legal needs), reach out to us at S2 Advertising Law - michael@lawofadvertising.com or wendy@lawofadvertising.com

AI (Artificial Intelligence) Makes Everything Easier in the Ad Industry. Right?

Remember way back at the end of 2022 when Microsoft-backed OpenAI’s new chatbot, ChatGPT, went public?  We were amazed at the human-like capabilities and creativity of this very non-human resource.  Want to hear a unique poem about flowers?  No problem.  Hankering for a philosophical discussion about the meaning of life?  Let’s do it.  Need help debugging or writing a piece of code?  ChatGPT was there for you.  The biggest thing we were concerned about was our kids finding out about it and never turning in their own homework ever again. While ChatGPT wasn’t the only (or necessarily the best) AI chatbot out there, Microsoft’s backing, and its affiliation with the search engine Bing, definitely brought this world out of the sci-fi genre and more into the mainstream than it had ever been before.

Notably, the advantages to the ad industry seemed clear right away.  Advertisers could type in a few terms or pointed requests, and create fancy artwork, “original” copy and solutions to problems.  

We quickly saw that AI may even be able to fully generate real ads.  In January 2023, Hardee’s asked the AI tool DALL-E to imagine Hardee’s SUPER STAR Burger, resulting in 10,000 crazy burger images and a big social campaign.  The thrust of the campaign was essentially, “don’t worry, AI still can’t create our burger,” but it certainly got marketers thinking about faster, cheaper solutions to ad content creation.  We also saw Ryan Reynolds ask ChatGPT to write the copy for a social spot for MintMobile.  While it possibly wasn’t the most creative ad, it got the point across.

You couldn’t blame advertisers for considering the potential cost savings if AI could effectively do the jobs of artists and copywriters.  

However, months into this “post-ChatGPT” world, some real concerns have started to emerge.  The NY Times’ Kevin Roose’s unsettling experience with ChatGPT on Valentine’s Day night 2023 freaked everyone out.  The chatbot’s (or “Sydney,” as it wanted to be called) insistent love, professed desire to hack into nuclear codes, and potential to sway vulnerable folks to perform bad acts, had Microsoft admitting there’s work to be done as it irons out the kinks.

And before hard-working creatives fall into a full-out panic about their job security, there are still a number of obstacles on the legal side that brands and agencies will need to work through:

• Who owns the output of AI machines?

An agency may input its own original thoughts and directions into an AI tool, but who owns what comes out of that?  Unfortunately, the law is not yet clear on this issue.  Who might own the output? The agency? (because it started the process); The chatbot? (because it created the ultimate work product); or maybe even the owner(s) of other content used by the chatbot to generate the ultimate product. So, if you need to own your resulting ad content, using AI may complicate matters. 

Further, if you use AI to create designs for your ad, it’s still unclear at this point whether you can protect those designs via copyright law.  Artwork need not necessarily have copyright protection to be usable in an ad, but being able to protect work via copyright affords an owner additional rights and protections (e.g., ability to sue in federal court, more easily obtain an injunction, seek statutory damages/attorneys’ fees, etc.).  In a case ongoing since 2019, the United States Copyright Office refused to register a copyright in an AI-generated piece of artwork, explaining that the Copyright Act’s requirement for an “original work of authorship” actually also requires “human authorship” to support a copyright claim. And art produced via AI did not qualify.  The artist using AI is challenging the decision.  However, in March 2023, the Copyright Office issued a statement, clarifying that works that contain AI may indeed be eligible for copyright protection, if the work maintains some human authorship (and then, copyright will only protect those elements that are human-authored).  It’s complicated!

• Can the AI machines freely take from anyone/anywhere on the “input” side?

What about on the other end of AI?  What if you don’t want all these AI tools using your hard-earned work to create new content?  If you, as an ad agency, brand or creative have released your marketing content to the internet, even if you could technically block AI machines from using your content as training data, can you legally stop it?   There are at least 3 ongoing cases tackling this issue.  The cases deal with the rights of AI machines to (a) ingest existing, copyrighted, source code to generate code for others (Doe v. GitHub, N.D. Cal. Nov 10, 2022); (b) use existing art to create art “in the style of” the original art (Andersen v. Stability AI et al, N.D. Cal. Jan. 13, 2023); and (c) take stock images without licenses to feed their reservoir of imagery (Getty v. Stable Diffusion, Jan. 2023, UK).

Advertisers may be struggling to come to terms with the fact that if they will benefit on the output side of AI, they may have to live with the loss of rights on the input side.  Or, will the courts severely limit the AI machines’ gathering capabilities before they even get going, making the tools less desirable?  Courts will struggle with what is the best policy outcome for the developing technology.

• Will anyone protect AI-using advertisers from claims?

And finally, what happens if you use AI for your ad, and then a third party comes along and claims you infringed their rights?  That fantastic AI machine created the ad, so it will stand behind it, right?  Of course, the answer is “not so fast.”  

What kind of claims are we talking about?  While they really could run the gamut, it’s possible that your AI machine could generate content that references celebrities, incorporates others’ trademarks, or even includes copy that is substantially similar to an existing campaign.  When you use that content, you run the risk of right of publicity, trademark and copyright claims from the celebs, trademark owners, and originators of the copy.  

If you turn to your AI provider and ask it to handle those claims, those asks are likely to fall on deaf ears.  Their terms do not promise unlimited commercial use.  They don’t promise all output has been cleared for ad campaigns.  And they certainly don’t promise to indemnify you if you get a claim.  Luckily for us ad lawyers, AI machines do not submit themselves to legal review before releasing the content.  

As we all know, advertising copy is afforded the least protection under the First Amendment, and what you may be able to get away with for pure entertainment or news purposes won’t fly when you use it to promote your business and to make $.  At this stage, these AI machines may be great at producing usable entertainment/news content, but when it comes to advertising, we reach an entirely different risk level.  For advertisers, it really is “use at your own risk.”

This is all not to say AI is unusable in the ad industry today.  There may be some unique, lower-risk tasks for which AI may be ideal.  For example, brands and agencies have seen a lot of success with AI-assisted search engine optimization (SEO) and audience targeting.

However, while the day may come that AI replaces us all, at least in the ad industry, we’ve still got some hurdles to overcome before that becomes a reality.  

__________

For legal questions about AI and the ad industry (or any advertising/media legal needs), reach out to us at S2 Advertising Law - michael@lawofadvertising.com