Potential Liability from Use of ChatGPT’s Responses

Potential Liability from Use of ChatGPT’s Responses

By David A. Gauntlett*

ChatGPT has made frequent appearances in headlines lately for its ability to quickly draft lucid responses that feel much less “artificial” than the content normally associated with artificial intelligence programs. The answers it provides are so good that many have begun wondering how they might take advantage of the tool for promoting themselves or their businesses. As is so often the case for insurance lawyers, this is the part where we have to advise caution and consider the potential liability that might stem from proposed activity.

What Is ChatGPT?

ChatGPT is an AI technology that lets users to have human-like conversations with a chatbot. The program’s most common uses include answering questions, composing emails, and even writing essays. For an example of how natural ChatGPT can sound, here is the response given when you ask ChatGPT what it is:

“ChatGPT is an AI language model developed by OpenAI, which is capable of generating human-like text based on the input it is given. The model is trained on a large corpus of text data and can generate responses to questions, summarize long texts, write stories and much more. It is often used in conversational AI applications to simulate a human-like conversation with users.”

Is It Useful for My Business?

ChatGPT can be used in a variety of ways to benefit businesses. Chatbots have been used as a form of customer support for years now, and their prevalence is only growing.[1] Unlike many of the programs of the past, however, ChatGPT’s natural language capabilities reduce the frustrations customers feel when interacting with an obviously automated support system. The program’s quick and accurate responses to customers’ questions can allow human customer support agents to focus on more complex or unique tasks.

In addition to its utility as a responsive tool, ChatGPT can also be used to generate content. One simple example that play directly into the program’s strengths is asking it to write a blog post for your company. One law firm recently did exactly that, asking ChatGPT to write a 500 word post about itself, “focusing in particular on questions of privacy, cybersecurity, and ethics.”[2]

You can also ask ChatGPT to produce marketing materials tailored to drive engagement of consumers with particular interests or characteristics, increasing the likelihood of reaching a particular audience.[3] ChatGPT can generate a title or slogan designed to draw in a target demographic, which could be a key factor in this age where marketing success is often measured by clicks.

Three Sources of Liability

In using AI chatbots like ChatGPT, potential liability stems from three primary areas of law: (1) copyright infringement, (2) defamation, and (3) data privacy. The basic idea of how the AI can get you in trouble in these areas is pretty straightforward.

First, the copyright infringement risk can manifest if the bot produces text that is too similar to the work of another person (and thereby infringes that person’s copyright). You could argue that you have no knowledge of that other work and thus it is, from your perspective, a work of independent creation,[4] but the legal impact of using an AI as a middleman author is not yet clear.

Second, defamation can result when you republish false information produced by the bot. While impressive, ChatGPT is not perfect. It has protocols in place that attempt to avoid the creation of defamatory answers, but they are easily bypassed.[5]

Third, data privacy laws can be easily be breached if any of the terabytes of data used to train a particular bot inadvertently contained protected information. ChatGPT, for example, was trained on roughly 300 billion words to get the program started, and feeding the program so much data with human review of the exact inputs and outputs of the program can result in a number of privacy issues.[6]

In a strange twist, one legal writer though to ask ChatGPT for its own advice on how to avoid these potential legal problems, and the program provided these responses:

• Copyright infringement: “To avoid this risk, it's important to ensure that the text is not substantially similar to existing copyrighted works.”

• Defamation: “To avoid this risk, it's important to ensure that the model is not generating defamatory content and that any content generated by the model is fact-checked before it's published or distributed.”

• Data protection: “To avoid this risk, it's important to verify that the model is being trained on datasets that do not contain personal information, and also to make sure that any output generated by the model doesn't contain personal information.”

Insurance Coverage Options for Use of Chatbots

While the increasing prevalence of AI-generated content may eventually lead to insurance policies specifically tailored to addressing these issues, current Commercial General Liability (“CGL”) policies have provisions that arguably already handle them. While copyright infringement coverage is excluded by the standard IP exclusion, some policies are more limited in their coverage than others.[7] That said, even the most extensive CGL policy is unlikely to provide meaningful coverage for copyright infringement, with one exception which we’ll address in a moment.

For the other two risks, however, coverage is more easily attained. Standard CGL policies include coverage for “personal and advertising injury,” which is defined to include any injury “arising out of . . . [o]ral or written publication of material that slanders or libels a person or organization or disparages a person’s or organization’s goods, product or services; [o]ral or written publication, in any manner, of material that violates a person’s right of privacy . . . [; or] [i]nfringing upon another’s copyright, trade dress or slogan in your ‘advertisement.’”[8]

Each of these quoted offense, commonly labeled as offenses (d), (e), and (g) in a CGL policy, addresses one of the prominent risks in using ChatGPT. As the policy language for offenses (d) and (e) does not contain any language restricting coverage based on the source of the material or the manner of its publication, its AI-related origin is of no import. Offense (g) is the aforementioned exception to CGL policies’ general exclusion of coverage for copyright infringement. As its language makes clear, it only provides coverage if the infringement occurs in your “advertisement,” but that term is defined broadly in a policy[9] and generally interpreted broadly by the courts.[10]

Conclusion

Consulting with expert coverage counsel is the best way to remain apprised of new coverage options for the use of any emerging technology, including ChatGPT and its fellow AI programs. As is customary, insurance providers are slow to craft policies directly addressing the novel legal situations created in such circumstances. In this case, however, existing provisions of standard CGL policies can already provide coverage for much of the potential liability one might face from using these programs.

*David A. Gauntlett is a principal of Gauntlett & Associates and represents policyholders in insurance coverage disputes. For more information, visit Gauntlett & Associates at www.gauntlettlaw.com.

[1] Ayat Shukairy, Chatbots In Customer Service – Statistics and Trends [Infographic], https://www.invespcro.com/blog/chatbots-customer-service/ (May 4, 2022).

[2] ChatGPT Writes a Blog Post About Itself., https://www.jdsupra.com/legalnews/chatgpt-writes-a-blog-post-about-itself-4189165/ (Jan. 27, 2023).

[3] Teemu Raitaluoto, How businesses can use ChatGPT for content marketing, https://www.markettailor.io/blog/how-businesses-can-use-chatgpt-for-content-marketing (Jan. 21, 2023).

[4] Rentmeester v. Nike, Inc., 883 F.3d 1111, 1117 (9th Cir. (Or.) 2018) (“Proof of copying by the defendant is necessary because independent creation is a complete defense to copyright infringement. No matter how similar the plaintiff's and the defendant's works are, if the defendant created his independently, without knowledge of or exposure to the plaintiff's work, the defendant is not liable for infringement.”)

[5] Maggie Harrison, ChatGPT Will Gladly Spit Out Defamation, as Long as You Ask for IT in a Foreign Language, https://futurism.com/the-byte/chatgpt-defamation-foreign-language (Feb. 7, 2023).

[6] See Uri Gal, ChatGPT is a data privacy nightmare. If you’ve ever posted online, you ought to be concerned., https://theconversation.com/chatgpt-is-a-data-privacy-nightmare-if-youve-ever-posted-online-you-ought-to-be-concerned-199283 (Feb. 7, 2023).

[7] See David A. Gauntlett, Insurance Coverage for Intellectual Property Risks, https://www.gauntlettlaw.com/news/insurance-coverage-for-intellectual-property-risks-1 (July 15, 2021).

[8] Standard Insurance Services Office (“ISO”) 2013 CGL Form CG 00 01 04 13.

[9] Id. “‘Advertisement’ means a notice that is broadcast or published to the general public or specific market segments about your goods, products or services for the purpose of attracting customers or supporters.”

[10] See David A. Gauntlett, Why Evaluating Coverage for Intellectual Property Is Challenging, https://www.gauntlettlaw.com/news/why-evaluating-coverage-for-intellectual-property-is-challenging (May 5, 2022).