OpenAI and Copyright
I know copyright infringers.
They tried to steal my company's website.
We stopped them with DMCA take-down notices.
OpenAI isn’t that kind of copyright infringer. Indeed, I have seen very little that suggests any text-based LLM, whether from OpenAI, Meta, Google, Microsoft, Hugging Face, Baidu or any of the others feels like copyright infringement.
I’ll explain.
Caveat: I am not an attorney nor have I been trained in the details of copyright law. But I have some first-hand experience.
With a partner, I started The Analyst Syndicate (TAS) in the fall of 2018. We were a robust collection of esteemed former Gartner analysts who had stepped back from working for Gartner, heading eventually for full-time retirement but who didn’t want to quite go from red-line efforts to lumpy idle and then shut down with a small post-ignition shake. (In future times when EVs dominate and few have experience with internal combustion engines, even fewer will get the metaphors in this paragraph but that time isn’t now.)
We grew TAS by leaps and bounds in the first year, with a good 40 analyst-members. We met weekly to discuss what was going on in the IT industry and the applications and business implications thereof, as well as specific projects we had taken on for various clients, papers we were working on, and business that we were pursuing.
We published about 100 papers on The Analyst Syndicate website (www.thansyn.com) which can still be found on The Wayback Machine. (I’ve included at the bottom of this note a list of pointers to some of our research our that site.)
Business was good in the first year, OK in the second year, slowed further in the third year, and, by the start of 2022, the handwriting was on the wall, floor, and ceiling. We just weren’t generating new business. Which, after fits and starts, we accepted. (That’s consistent with what friends in other industries experienced when they hung out a shingle near the end of their careers to stave off the impact of cold-turkey retirement.)
We shut down TAS in August of 2022. (Most of us had our own personal business entities, LLCs, and the like, which many — but not all — shut down that fall, a year ago. Not all have fully retired, however, and a few are going great, still generating revenue and growing new business.)
Fast forward to this spring. A good friend who worked with me at Gartner had just retired from there and we got on Zoom to catch up. She wanted to know how much I had gotten for the TAS website. That led me to discover someone else had bought limited rights to the URL (www.thansyn.com) and was in the process of (a) rebranding and repositioning it as “Tech Horizons” and (b) posting all our old research back up on their new website with new authors’ names.
The spirit of The Analyst Syndicate sprang back to life for the next three weeks as we found their hosting site and sent the firm hosting their website DMCA takedown letters for 59 of the documents they had purloined, along with several other copyright violations. Here’s an explanation of takedown letters and a deeper dive into the Digital Millenium Copyright Law.
The end result? They were thrown off their hosting site and have not reappeared as far as I know. They’re gone. Why? They clearly violated the intent of copyright law.
Shelly Palmer’s blog post today wrote a good summary of OpenAI’s defense against claims that ChatGPT violates the copywrites of millions of authors whose work was fed into ChatGPT to train its Large Language Model. Alternatively, you can also read OpenAI’s motion to dismiss the copywrite lawsuit.
Their defense considers commercial intent and legal precedents on fair use, publishing statistical information like word frequencies, consideration of what constitutes a derivative work, and how products like book reviews don’t violate the original author’s copyright.
There’s a complex collection of issues here. At 38 pages, the entire motion to dismiss deserves a read.
What do you think?
Here’s a list of some of what The Analyst Syndicate published. Fifteen articles. Access is free. Don’t change the authors’ names! And don’t violate our copyrights!
(If you want a much longer list, let me know. I can mail it to you.)
Addendum
All of these papers are hosted on The Web Archive which is slower than a commercial site. It’s a free site, great for preserving material without having to spend money on a commercial site. So have some patience as you want for the site to get you to the right page!