Blog

4 Things Businesses Should Know About ChatGPT

Workers in your business are probably trying out ChatGPT. Here are 4 things you should know.

#1 – You don’t own what it writes. ChatGPT output is not copyrightable. The Copyright Office explains: “When an AI technology determines the expressive elements of its output, the generated material is not the product of human authorship. As a result, that material is not protected by copyright.” So, if your competitor copies your ChatGPT generated content, there won’t be much you can do about it. ChatGPT may even help them do it. From OpenAI: “[d]ue to the nature of machine learning, Output may not be unique across users and the Services may generate the same or similar output for OpenAI or a third party.”

#2 – You may need another’s permission to use what it writes. ChatGPT “learns” by ingesting material that it can reproduce, at least in part, in response to user prompts. As claimed in the mounting number of lawsuits targeting generative AI, the material ingested can include copyrighted material. Therefore, the output could include that copyrighted material, and your use of it could be copyright infringement. OpenAI grants users rights in the content generated by ChatGPT, but that won’t prevent a copyright owner from suing you.

#3 – It can’t keep a secret. Whatever information you input is not confidential. ChatGPT is ingesting and storing all the information it receives, and using it to improve its answers to prompts. So, its very purpose is to use your information to inform others. From the opening screens: “Conversations may be reviewed by our AI trainers to improve our systems…Please don’t share any sensitive information in your conversations.” There is an option to prohibit use of your data to train the AI, but exercise of that option requires trust in both your employees to use it and OpenAI systems to actually honor it.

#4 – It makes stuff up. You need to fact-check ChatGPT output. This right from opening screens: “the system may occasionally generate incorrect or misleading information….” From the Your Rights Section of the OpenAI Privacy Policy: “you should not rely on the factual accuracy of output from our models.” Lawyers will know the story of the attorney who used ChatGPT to write a brief that included citations to cases that did not exist. Don’t be that guy.

Warhol v. Goldsmith – My Hot Take

Some interesting and telling moments in yesterday’s Supreme Court oral argument of the Warhol v. Goldsmith case started with the Syracuse hypothetical Justice Thomas posed to the attorney for the Warhol Estate. At bottom the case is about whether Warhol’s orange color alteration of Goldsmith’s photo of Prince qualifies as copyright fair use. Justice Thomas asked: “let’s say that I’m also a Syracuse fan and I decide to make one of those big blowup posters of Orange Prince and change the colors a little bit around the edges and put ‘Go Orange’ underneath. Would you sue me?” Warhol’s attorney hypothesized that if Warhol did sue for that, Warhol should win even though Goldsmith should lose this case, basically because Warhol’s work is more creative than Justice Thomas’.

I can’t see why, or that it should matter. Warhol’s argument is that Orange Prince is fair use because Warhol “transformed” the “meaning” of Goldsmith’s photo of Prince. According to Warhol, the photo show’s Prince’s vulnerability but Warhol’s addition of color and other alterations render Prince in a more “iconic” way. If Warhol is correct, Thomas’ poster must qualify as fair use of Warhol because it “transforms” Warhol’s work into one demonstrating support for the ‘Cuse. The Warhol Estate’s unwillingness to concede the Syracuse hypothetical seems unprincipled and likely to hurt its case. Justice Thomas’ hypothetical highlights that one cannot expand the scope of fair use without diminishing the scope of copyright. If the Warhol Estate doesn’t have to license Orange Prince, its own copyright protection must be similarly narrow and Justice Thomas shouldn’t have to license Go Orange. Which way should it be?

U.S. Copyright Deposits Are Important

U.S. copyright applications must be accompanied by a deposit of “complete copies of the best edition.” Two recent cases underscore the importance of deposit copies. In Pyrotechnics Management v. XFX Pyrotechnics, plaintiff claimed defendant infringed by copying computer commands designed to manage firework displays, generated in accordance with a method protocol. The deposit copy, however, included only a description of the protocol (basically a manual), only three of the actual messages generated in accordance with the protocol, and no code. Therefore, the court held that plaintiff’s copyright did not include code. It was limited to the description of the protocol, and not even the protocol itself because the protocol itself was an uncopyrightable method. Because defendant did not copy the description of the protocol, plaintiff lost. Had the plaintiff deposited code perhaps the result might have been different.

In Parker v. Hinton, No. 3:19-CV-00214, 2022 WL 706920 (M.D. Tenn. Mar. 8, 2022), plaintiff was unable to prove that he deposited any material at all. Guess what. He lost too.

Bottom line: the deposit copies matter. They are required, and they stake the outer boundaries of copyright protection. The deposited copies should be a complete and accurate copy of the copyrighted work. The deposit copy should be the first piece of evidence examined by both plaintiff and defense counsel in litigation.

Connecticut Personal Data Privacy Law – What to Know

Connecticut passed a personal data privacy law this year. Here’s what businesses should know.

  • The law includes industry-specific and data-specific requirements, so business-specific analysis is required.
  • The law takes effect July 1, 2023, so there is time to prepare.
  • The law applies only to businesses possessing data of 100,000 or more CT consumers (if the business earns more than 25% of revenue from the sale of consumer data the number is 25,000). Non-profits, institutions of higher education and certain financial businesses subject to other regulations are exempt. HIPAA and other medical information, credit information, and certain other categories of information are exempt.
  • Businesses must disclose: the categories of personal data they process and the purpose of the processing; whether the business shares or sells personal data; and how consumers can access, modify, delete, or opt-out of use of their data.
  • Businesses must allow consumers to access their data, request copies, correct inaccuracies, and delete their personal data, and must comply with requests within a reasonable time, presumptively 45 days.
  • Businesses must employ “reasonable” security practices to protect consumer data and document those measures with respect to data that could be especially harmful to consumers if disclosed. That includes data used for targeted advertising or profiling.
  • Use of consumer data to improve functionality is permitted.
  • More strict requirements apply to child data.
  • There is no private right of action for violation of the law. Only the Connecticut Attorney General can enforce it. That means businesses will not be flooded with class action and contingent fee lawsuits on day one, and until 2025 the AG is required to give 60-day notice of any violation and an opportunity for the business to cure it before bringing suit.
  • Violation of the law is a violation of CUTPA, the Connecticut Unfair Trade Practices Act, for which remedies include punitive damages and attorneys’ fees.
  • Connecticut joins California, Nevada, Maine (limited to ISPs), Colorado, Utah and Virginia in enacting specific data privacy laws. The California, Nevada and Maine laws already are in effect. The rest take effect beginning 2023. Many of the laws are similar, but not identical, so if you have customers in all those states, you should know all those laws.

You can read the Connecticut law here.

Substantial Similarity 2022 Released

Release #19 of Substantial Similarity in Copyright Law is now available from PLI.

August 6, 2021

2021 Substantial Similarity in Copyright Law Update Released

Release #18 of our treatise Substantial Similarity in Copyright Law is now available from PLI.

Breaks for Trademark and Copyright Owners

The USPTO is allowing extensions for trademark owners affected by the Corona virus. The Copyright Office is basically closed for personal contact and mail, but moving along on all things electronic.

2020 Substantial Similarity Update

Last week my father and I finished the 17th annual update of Substantial Similarity in Copyright Law. As was the case last year, the 9th Circuit upset the apple cart near the end, this time issuing the en banc decision in the Led Zeppelin case just days before the update was due. We got it in there though, so look for the update in May. Hopefully by then, copyright will matter again. Be well.

Entertainment and Tech in NYC

On Thursday, September 26, Eric Osterberg will be speaking on a panel with Arthur Levy, Judith Finell and Robert Clarida about music copyright infringement cases at the Music Business Association’s Entertainment and Technology Law Conference in New York City. 

Busy Few Months

It’s been a busy few months writing and speaking. We finished the update to Substantial Similarity in Copyright Law in May, updated the Practical Law Substantial Similarity piece in June and taught AIMP Summer School in July. Thank you to all the clients who assigned work that was equally interesting and challenging this spring and summer.