Sometime last year, a client showed me some language that ChatGPT had generated for one of its employees who had turned to the tool to help write new language for the company website.  The language that ChatGPT generated was very familiar to me; it looked like I had written it, and I may have written it.  Previously, I had helped craft an argument to overcome a patent examiner’s objection.  Per our patent attorney, it is likely that ChatGPT had ‘scraped’ publicly available data from the U.S. Patent Office.  Since the patent was granted, the associated file wrapper became public and entered the public domain.

However, the information that the employee input (the “Input”) into ChatGPT would have also entered the public domain.  That is because generative AI systems utilize customer inputs as a source of real-world data to continuously improve their AI models.  Your Input can become someone else’s output.  (ChatGPT permits you to opt out of allowing your Input further ‘train’ its generative AI software.)

For business owners with trade secrets to protect, if your generative AI Input includes a trade secret, you may have failed to keep it secret.  And, if it is not secret, it is not a trade secret any longer.  Further, any NDA purporting to protect that information will cease to do so because it is now publicly available.

One painful example of this happening is discussed in an article by Forbes in May 2023.  Samsung Electronics banned the use of generative AI tools by its employees after sensitive internal source code was uploaded onto ChatGPT by an engineer the previous month.  Many large businesses have adopted similar bans.

There is now a big push for lawyers to use generative AI.  Articles, seminars, and webinars abound.  I have attended a few over the past year.  For the most part, the webinars and articles fail to inform lawyers about how the generative AI tools may put to use the information that the lawyers input  -- even an article in the Texas Bar Journal.  This is a significant omission because, for lawyers, if you input information that is client confidential information, you have breached your client’s confidence in violation of the rules of ethics. 

My Take:   It is possible to deploy generative AI tools in the workplace in a manner that protects the confidentiality of the Input information.  Unless you have deployed such a system, it is best to use these tools carefully, if at all.

May you find joy in what you do and who you are with.