Why Businesses Should Approach ChatGPT with Caution
One of the most popular discussions in technology conferences around the globe at the moment is artificial intelligence, specifically OpenAI’s ChatGPT tool. If you have not used it yet, ChatGPT is tool that allows a user to input a natural language question and get a reasonably coherent answer.
Is that answer correct? Maybe.
Should people begin using it in business as a substitute for human-created content? Not yet. Here are the reasons businesses should approach ChatGPT with caution for the moment. For those businesses that are advancing their objectives with ChatGPT, it is imperative that they have well-developed policies to guide their employees and reduce the risks related to AI usage.
A Constantly Evolving Knowledge Base
The first reason businesses should not rely on ChatGPT with mission-critical issues is because OpenAI says it is not ready for such heavy lifting. While the knowledge base is growing, the reliability of the output depends on whether there is a good base of knowledge in the particular subject. Users will not be sure whether the answer is based on data that has been subjected to statistical validation, or whether it will run afoul of privacy protections. Even when the answer is reliable, the output may not be appropriate given the context. And, if the answer is wrong or harmful, ChatGPT disclaims all legal liability for using it.
The next reason businesses should be leery of using ChatGPT is the licensing implications. ChatGPT’s terms of use clearly indicate that while a user may own any input it provides, ChatGPT can use that input for machine learning, to improve its services, and for a host of other purposes. Furthermore, if other users input similar models (questions or queries), those users may get identical output.
ChatGPT and The Copyright Act
If a business wanted to use ChatGPT to generate marketing content, it is arguable that the content would not be protectable under the Copyright Act because the chatbot is not an author, and the output may not be unique. Anyone who inserted a similar model could receive similar or identical output. Universities are instituting policies that use of chatbots is prohibited and will be considered plagiarism.
Issues involving ownership of the input, the machine learning, the fine-tuning, the output and everything in between have not even begun to be meaningfully litigated, and battles are likely to be lengthy and bitter.
ChatGPT claims it has 100 million monthly subscribers, and that it is growing exponentially. That’s a lot of potential arguments over the ownership and usage of data. When you add that to the belief that AI-generated output is notoriously unreliable, the situation becomes fraught with peril.
OpenAI Rapidly Changing Terms of Use
Finally, the OpenAI terms are rapidly changing. There was one set of terms before February 21, 2023, another set of terms published on February 21, 2023, and yet another set of terms published on March 1, 2023. It is difficult to keep up with the many licensing and terms of use changes. The current version of the terms does indicate that API users’ content will not be used to improve OpenAI’s services, and OpenAI reiterates that now users own their own content, but the terms of use do not include any language about who owns the machine learning or the output. OpenAI does currently say that it will not use content provided by users to train or improve its models, though, without the users’ permission. The in-flux nature of the terms is also a cause for concern.
Despite these concerns, many businesses are starting to use the technology to develop content. If your business is one of those that plans to tread carefully into the early stages of ChatGPT, it is critical to develop and publish a policy that clearly articulates the parameters of what is acceptable and what the expectations are with respect to safety and ownership. This can be difficult, even if your business does not plan to use ChatGPT until some of these issues are more settled. For instance, Walmart initially had an outright ban on using chatbots and related technology. Last week, Walmart removed the ban and instead asked employees to refrain from inputting any sensitive information, shopper information, or confidential knowledge into ChatGPT or other AI portal.
If your business is planning to use ChatGPT’s API, it is important to understand the intellectual property and licensing implications, as well as the impact it may have on your need to obtain revised agreements with your clients.
If your business needs assistance developing, revising, or evaluating its policies related to ChatGPT or other AI technology, please contact me at julie@scottiplaw.com for a confidential discussion about next steps.
