Right now, the hype around AI continues to boom while the evolution of the tools keeps accelerating. The massive leaps in capabilities in the last 24 months have gone from the much-joked-about "AI wrote me a script" memes to strikes and walkouts over the AI usage in the film and television industry. The technology keeps progressing and breaking new frontiers and regulations and governance over its usage are lagging behind.
Many talk about what the next big leap in AI tech capability will be, or discuss the ethical implications of taking human creativity out of art, but an area that I'm interested in is where is the law going to go around all of this technology. Copyright infringement, corporate liabilities, slander and misrepresentation... there are so many ways in which this tech needs to be regulated.
How are governments and corporations going to agree on the rules on how this will be governed? How will people be held responsible? I recently said on a podcast: "I can't imagine what would happen if Disney found out people were just generating Marvel stuff and making Marvel movies and distributing it and they have no rules around how to control that."
I think back to the digital music revolution with Napster. There was a culminating point in time where we had progressed far enough in technology to be able to create digital copies of music at a high enough quality that people wanted it. We also had solid enough internet bandwidth to share that music. And tech was able to advance and make it possible for regular people to connect with each other and share their libraries. The music industry lost its mind over this! They weren't in the middle any more. Artists were making music, and it was getting out to people, and the corporations couldn't control the methods of distribution. It took several years for the courts to catch up on how to make the rules around that and today we've all moved to paying to download music from a library, or listening to a stream for free, but artists get their royalties again and the corporations are back in control.
Recently, many videos on TikTok had been silenced because copyrighted songs were pulled from legal usage. TikTok only very recently came to a new licensing agreement with Universal Music Group, restoring the ability for creators to use the songs in their catalog.
However, we haven't reached that point with AI technology yet, especially generative AI tech. Many legal organizations are making their first decisions in this area, setting precedents for the future. Jacqueline Baxter recently spoke about a new bill addressing the use of copyrighted materials in generative AI models. She and I chatted about this topic and the coming shifts in legislation focus.
"Schiff's bill would not ban AI from training on copyrighted material, but would put a sizable onus on companies to list the massive swath of works that they use to build tools like ChatGPT -- data that is usually kept private."
**Source: https://www.theguardian.com/technology/2024/apr/09/artificial-intelligence-bill-copyright-art
Jacqueline Baxter
Director, Experience Design & Content Strategy
If you are using generative AI within your organization (and I'd be willing to bet that's most of us at this point), you should have eyes on the developing conversation around AI regulation. "Is this legal?" is a massive question, and it will impact the whole AI industry. The ripple effects of these decisions will include all the companies and individuals using AI tools. There are plenty of confident people on both sides of the regulation argument, but the truth is...we don't know how any of this will shake out.
Understanding how models are trained, where the training data comes from, and the potential ethical and legal concerns of using these tools is the responsibility of every company using generative AI technologies. When in doubt, watch your six.
Jason St-Cyr
Director, Developer Relations
I see this being the real new innovation sphere in AI: legal use and compliance. So far, it's been fun and open and untamed, but this is the logical next step. Big corporations are going to want to protect their IP.
Baxter
That's exactly it; how do we create AI systems that benefit some industries without, at best, disenfranchising or, at worst, actively harming others. I can easily see a scenario where it's ruled legal to use copyrighted data to TRAIN a model but any OUTPUT of that model could be subject to copyright infringement.
St-Cyr
Ah, but what if the output is nothing like the copyrighted data it was trained on? A musician who listened to a bunch of different artists and then created something new that is unlike their influences.. are they infringing on copyrights? For example, what if you learned Bach and Mozart and the Beatles... and then wrote Piano Man. Is Piano Man an infringement because Billy Joel listened to Hey Jude? The music industry started building really specific rules around this that relied on analyzing the output and its similarities to the inputs. Substantial reuse became a huge piece of determining if something was a copyright infringement.
I think that down the road we're going to get to that point where the output is analyzed for a specific percentage of reuse of an intellectual property in the output.
Baxter
I think that's right -- at least a few of the pending lawsuits are based on the claim that generative AI technologies are prone to verbatim recreation of text that's under copyright. When it comes to written text, even paraphrasing without giving credit can be considered plagiarism -- that's true even if you're quoting your own work! Plagiarism isn't a legal standard, but it might be useful for us to ask ourselves about plagiarism (which is more of an ethical question) rather than copyright (which has a legal definition) when it comes to brand use of these tools. The technology is moving in this direction as well -- Copilot has started providing footnotes when providing specific content!
If the technology is genuinely producing new copy, that opens a question about to whether it's eligible to be under copyright at all. A US judge has said that AI-generated art is not eligible for copyright since it does not contain human creativity. On the other hand,The US copyright office just granted a copyright on a self-published novel that extensively used ChatGPT where the copyright holder is considered the author of the "selection, coordination, and arrangement of text generated by artificial intelligence" which is an interesting approach since (as far as I can tell) that means the paragraphs and sentences aren't protected but the work as a whole is subject to some protections.
"*Matthew Sag, a professor of law and artificial intelligence at Emory University, calls what the USCO granted Shupe "thin copyright"---protection against full-fledged duplication of materials that doesn't stop someone from rearranging the paragraphs into a different story. "This is the same kind of copyright you would get in an anthology of poetry that you didn't write," Sag says. *"
St-Cyr
That's interesting, I hadn't seen that one! I suppose it's not unlike somebody taking news clippings and pasting them together in a collage. You can't own the copyright to the text from the news clippings, but the arrangement of them into a form of art could be copyrighted. There is going to be such a messy few years on this!
Baxter
I think that's exactly right -- the complete work has been judged worthy of copyright, but the individual pieces weren't. Hard to say how far that will reach; most of these seem to be taken on a case-by-case basis. At some point there are going to be more far-reaching laws and cases. There are several lawsuits pending that address questions of copyright; it's a whole new world out there! I think the important thing is that brands using generative AI tools of any kind keep an eye on the evolving legislation and be prepared to pivot to accommodate changes to the environment.
Credits
- Content: Written and edited by Jacqueline Baxter and Jason St-Cyr using good ol' fashioned human writing.
- Cover image: Generated by Jason St-Cyr using NightCafe
- Title: Generated by Jason St-Cyr using ChatGPT
Top comments (0)