AI is shiny. AI is new. AI is sexy. And AI is problematic and not entirely understood. It is for those last reasons that, according to HR Brew, corporate America is either restricting employees' use of generative AI tools such as ChatGPT or banning them outright.
I don't have the answer for which of allowing, limiting, or banning generative AI is the correct answer for your business. You should, however, consider these three risks in evaluating whether, when, and how your employees use generative AI at work.
1.) Confidentiality and Privacy. Employees could share trade secrets or other confidential information through conversations with generative AI. While terms of service make it clear that AI services don't retain user-supplied information, they do learn from it. Thus, the possibility exists that the AI will inadvertently misappropriate trade secrets, et al., and re-supply to others in a future conversation. Thus, employees should be cautioned not to supply and trade secrets or confidential information through AI chats.
2.) Accuracy. When I asked ChatGPT, "Tell about employment lawyer Jon Hyman," it told me that I am the founder of the Hyman Law Firm (wrong), and that I attended Ohio State for law school (also wrong). It did get a lot of things correct, but I could easily answer those two simple questions with a quick Google search. I've even seen ChatGPT and other generative AIs create sources and facts out of thin air. Your employees need to understand that generative AI can be acceptable starting point for research, but it can never be the end point. Every answer it provides, every kernel of information it offers, must be vetted and cross-checked for accuracy. It will get more correct than it gets wrong, but it's the wrong ones that will kill you.
3.) Discrimination. If generative AI relies on biased information, it could supply biased answers to chat prompts. Garbage in, garbage out. The law considers disparate impact protected class discrimination (unintentional discrimination based on something that tends of favor one class over another) just as illegal as disparate treatment discrimination (intentional discrimination). Thus, employees and their employers must vet any information provided by generative AI for inherent and other biases before relying upon them to make an employment decision.
Generative AI isn't going anywhere. Today's versions are very much alpha releases. Each subsequent release gets better and better. You can only hold off addressing its impact on your business for so long. Why not address it today and update your policies and internal guidelines as the AI becomes advanced. After all, if your employees are using AI to help them perform their jobs shouldn't you be providing guidelines for its use and guardrails around its misuse?