Copyright | Notion

Recent

I’ve been watching the intersection of entertainment/media and genAI lately. It’s a hot spot, and a testing ground for where the legalities of all this might be headed.

https://www.innovationleader.com/moves-that-matter/hollywood-studio-lionsgate-strikes-data-deal-with-genai-startup-runway/

https://carey.jhu.edu/articles/navigating-legal-frontier-ai-hollywood

</aside>

And…

<aside> 💡 GenAI models are trained on vast amounts of public data, which often include copyrighted works. This immediately raises concerns about potential copyright infringement in both the training process and the generated outputs.

Recent example:

After the May 20th 2024 public demo and rollout of chatGPT 4o by OpenAI, the actress Scarlett Johannsen threatened legal action for the system’s avatar “Sky”’s uncanny likeness to her own. Read more:

What the current status of the dust up between Scarlett Johannsen and OpenAI,...

In terms of who owns AI generated works, US Copyright law does not recognize your ownership of a work created by genAI when you go straight from prompt text to brand new output. In practical terms, say your a musical group making a new album and use Midjourney for the cover art. You do not have an exclusive use of that image. Other bands, or anyone, can re-use your album cover at will.

Exceptions:

Original copyright adheres when a work is only modified, e.g., converting a photo from color to black/white. Or applying filters.
generating test code for a piece of code you wrote is likely also protected.
If the output is only a small part of a larger human-made work, you’re covered.
Of you used an LLM for a first draft, but then re-wrote it in your own words

#Promptography, a new term of art, and a controversy in the photography world, which gets very much to the heart of the issues here.

Sony World Photography Awards 2023

</aside>

Mitigation Strategies

<aside> 💡 Curtis’ recommendations:

FIRST, for IP and data readiness, highly recommend these two pieces from the World Intellectual Property Organization (WIPO). The first is “Generative AI: Navigating Intellectual Property,” and it addresses very specific mitigations your organization can take for the kind of risks I enumerated above. The second is, “Getting the Innovation Ecosystem Ready For AI: An IP Toolkit.” It advocates for taking proactive measures to adapt your own IP ecosystem for this new age of AI.; things like, updating your legal frameworks, providing clear guidance on IP protection for all your AI inventions, and considering the unique challenges posed by AI as an inventor itself.

AND, check out the work of the folks at National Institute of Standards and Technology (NIST) and their Risk Management Framework, ****or RMF. NIST’s Trustworthy and Responsible AI Resource Center created the RMF and updates it regularly, so I recommend you bookmark their site, or subscribe to their email updates. It’s dense, but not too technical, so it’s consumable by your full management team. It might help frame your conversations as you put AI governance in place. They even provide a “Playbook,” available as PDF or Excel spreadsheet, which is great to distribute at board meetings and speak with a common language.

</aside>

Specific Risks

Training Data Legality: Ensuring that the AI models they use are trained on properly licensed or public domain works to avoid infringement claims[2].
Output Infringement: The possibility of AI-generated content inadvertently reproducing copyrighted material, leading to infringement issues[4].
Fair Use Challenges: The ongoing debate about whether using copyrighted works to train AI models falls under fair use, which is being challenged in courts[4]. (see below “legal challenges under way”)

Ownership and Attribution Issues

There is significant uncertainty surrounding the ownership of AI-generated works[1]. Businesses must consider:

Copyright Eligibility: AI-generated works may not be eligible for copyright protection in some jurisdictions, as they lack human authorship[2].
Disclosure Requirements: The need to disclose AI involvement in copyright applications, as failure to do so can result in revocation of copyright registration[4].

Publicity Rights Concerns

The use of GenAI can potentially infringe on individuals' publicity rights, especially when generating content that includes celebrity likenesses or voices[4]. Key considerations include:

Unauthorized Use: Avoiding the commercial exploitation of individuals' names, voices, photographs, or likenesses without permission[4].
State-Specific Laws: Navigating the complex patchwork of state-specific publicity rights laws[4].