“Shall we play a game?” Those innocuous words “spoken” by Matthew Broderick’s computer in John Badham’s sci-fi techno-thriller War Games stunned audiences at the time. A computer that could “talk” and “think” and engage in conversation?!? This was the height of science fiction. Well, with the recent release of generative artificial intelligence (AI) tools, specifically in the form of ChatGPT and other predictive natural language processing (NLP) algorithms, science fiction has once again become reality.

Companies from Microsoft to Google and Instacart to Kayak have begun to incorporate and build upon this technology, originally developed by OpenAI. These tools can be incredibly beneficial to businesses, but they also carry risks.

Before we dive in to how generative AI can assist brands and companies, let’s first peel back the layers and understand – at a basic level – what ChatGPT is.

ChatGPT stands for Chat Generative Pre-trained Transformer. Let’s break that down:

• Chat refers to the interface that allows for interaction with the model using natural language prompts.
• Generative refers to a category of AI model that produces new output based on a given input. In practice this means that the “input” of a user query can generate the “output” of text, images, and audio answers.
• Pre-trained refers to the fact that the model has already been trained on a vast data set to teach it to predict the next word in a given sequence.
• Transformer refers to the architecture of the neural network (machine learning algorithms) upon which ChatGPT is based. It is this architecture that allows the computer to process natural language.

Importantly, ChatGPT and other generative AI are not omniscient; they cannot think, understand, or feel. They are merely software – lines of computer code – programmed to generate natural language replies in response to text and image prompts. They work by predicting the next word in a given text string based on patterns “learned” from the data on which they have been trained.

In late March, OpenAI released an API (application programming interface) to select businesses to allow them to incorporate the AI technology into their own websites and apps via plugins. Using these plugins, brands have been able to harness the power of ChatGPT to help consumers book travel, make restaurant reservations, and create curated product recommendations.

Currently, the ChatGPT plugins have been tasked with relatively basic functions – essentially providing high-level search tools in the form of an interactive chatbot. See, for example, the video Expedia released on Twitter to show how its ChatGPT plugin operates.

Brands have also begun to use generative AI tools to help with:

• Coding – generating and building source code and analyzing mistakes within the code
• Content Creation – generating blog posts, social media posts, targeted email campaigns and video scripts
• Data Analysis – analyzing large data sets and synthesizing the information into easily digestible bullet points
• Market Research – generating a list of key players in any industry along with products and services
• Product Descriptions – generating bulk descriptions for e-commerce sites where product catalogs are frequently updated
• Search Engine Optimization (SEO) – generating copy that includes keywords and meta descriptions that search engines can look for when ranking pages

ChatGPT and other generative AI tools come with risks, and any business use of them should be done carefully and cautiously.

First, generative AI is only as good and accurate as the information on which it was trained. Although vast amounts of data have been used to train the ChatGPT models – approximately 570 GB –  outputs “may be inaccurate, untruthful, and otherwise misleading at times.” [1] They can also “occasionally produce incorrect answers … harmful instructions or biased content.”[2] Further, ChatGPT “is confident and wrong a significant fraction of the time,” and the tool is “good enough at some things to create a misleading impression of greatness.”[3] In other words, don’t rush to replace your human customer service team with a generative AI bot, and make sure to have constant human oversight of any responses that the AI produces.

Second, there may be copyright risks attached to the use of outputs from generative AI in public-facing works. Because of how the ChatGPT model is trained, it is possible that responsive outputs could contain elements of copyrighted material. A brand that uses material generated by AI in a public-facing manner – such as in an ad or a social media post – could run the risk of infringing on another’s rights, for which the brand could be held liable. Accordingly, while these tools can be used as a launchpad for ideas, they should not yet be used to produce a final deliverable.

Additionally, under current U.S. intellectual property laws, copyright protection only exists where there is human selection and arrangement of information; there is no copyright protection for material generated by a non-human.[4] However, recent guidance suggests that for works containing AI-generated material, the Copyright Office will “consider whether the AI contributions are the result of mechanical reproduction” or “an author’s own original mental conception” to which the author “gave visible form.” Therefore, it is unclear whether material produced by generative AI without significant human input can even receive copyright protection.

Third, there are potential confidentiality issues to consider. Under U.S. law, to receive trade secret protection, a company must make a reasonable effort to conceal that information from the public. Allowing a third party such as OpenAI to have access to otherwise protected information without a confidentiality agreement in place could constitute a “public disclosure” and risk the loss of protection for the disclosed information.[5]

In addition to the above, there are privacy, security, ethical, and other legal risks to consider when using this technology or allowing any third party to access internal customer records and business data. For more information on these issues, read our sister blogs here, here, and here.

In the words of Stan Lee, “With great power, comes great responsibility.” New technology can innovate and help improve productivity and customer service. So go forward – but tread lightly. And before implementing any of these tools, consult with counsel to obtain an assessment of what legal risks could be triggered by their use.

[1] OpenAI, “What is ChatGPT?,” https://help.openai.com/en/articles/6783457-chatgpt-general-faq (last visited Apr. 10, 2023).

[2] Id.

[3] Berber Jin and Miles Kruppa, “The Backstory of ChatGPT Creator OpenAI,” The Wall Street Journal (Dec. 18, 2022, 2:57pm ET), https://www.wsj.com/articles/chatgpt-creator-openai-pushes-new-strategy-to-gain-artificial-intelligence-edge-11671378475, and Karen Hao, “What Is ChatGPT? What to Know About the AI Chatbot,” The Wall Street Journal (March 22, 2023), https://www.wsj.com/articles/chatgpt-ai-chatbot-app-explained-11675865177.

[4] U.S. Copyright Office, Compendium of U.S. Copyright Office Practices § 313.2 (3d ed. 2021).See also, Urantia Found. v. Kristen Maaherra, 114 F.3d 955, 957-59 (9th Cir. 1997) (holding that “some element of human creativity must have occurred in order for the [b]ook to be copyrightable” because “it is not creations of divine beings that the copyright laws were intended to protect”).

[5] According to OpenAI, any information provided to the services may be used to “provide and maintain the services,” and OpenAI may “review conversations to improve [its] systems.” https://openai.com/policies/terms-of-use and https://help.openai.com/en/articles/6783457-chatgpt-general-faq.

[View source.]