Artificial Intelligence

IP Implications of Generative AI

On March 9, 2023, Cooley presented a program in our New York office and via Zoom covering a new frontier in intellectual property protection and risk: generative artificial intelligence. Our lawyers discussed potential IP issues implicated by AI input and output, including copyright authorship, infringement and fair use. They also explored implications of AI for other areas of law, including trademark, trade secret, right of publicity, scraping and defamation.

Below are key takeaways from the recorded program. If you’d like to request a copy of the recording, please contact Jess Karas.

Key takeaways

Input concerns

• Copyright
  • Many generative AI tools are trained on publicly available data from the internet that may include protected works.
  • Not all unauthorized uses of copyrighted works are prohibited.
  • Fair use balances exclusivity against free expression and innovation.

Other input concerns

• The Digital Millennium Copyright Act/removal or alternation of Copyright Management Information.
• Scraping/unauthorized data access – e.g., contract breach, Computer Fraud and Abuse Act, trade secrets.
• Privacy concerns.

Output concerns

• Copyright
  • Is there substantial similarity in protectable expression?
  • How to separate similarity of unprotectable ideas and styles from protectable expression?
  • Is it fair use?
  • Who is liable for infringing output?
  • Infringement can be direct, contributory or vicarious.
• Right of publicity
  • Tools make it easy for anyone to copy appearance or emulate voice (deepfakes).
  • Right of publicity applies to commercial use.
  • Section 230 typically protects platforms from right of publicity and defamation claims but may differ by jurisdiction.

Other output concerns

• Trademarks
  • Expressive works receive First Amendment deference.
  • Will AI be used to develop new marks or assess conflicts/likelihood of confusion?
  • What if the output bears a trademark or logo that is likely to cause consumer confusion as to source or sponsorship – or that is likely to cause dilution of a famous mark (e.g., Getty v. Stability AI case)?
• Trade secrets
  • Prompts and output may be maintained (and disclosed) by third-party platforms, potentially negating protectability as a trade secret.
• Platform terms of use
  • May grant broad license to use of content.
  • “Assignment” of rights to user may be ineffective (17 USC § 204).
  • Subject to unilateral changes.
• Other risks
  • Disinformation, bias, and implications for ethics and professional standards.

What to look for next?

• The US Copyright Office is expected to release new guidance for generative AI soon.
• For now, users of generative AI platforms should be mindful that there is a risk that such works will be deemed unregistrable by the Copyright Office, at least in part.
• Companies and individuals wanting to avail themselves of copyright protection in works created with the aid of AI platforms should consider documenting the steps undertaken to dictate the specific contents of the work, including the extent of any expressive contributions added to AI-assisted output. 

Recent decisions and other resources providing more insight on this evolving area of law

• Thaler v. Perlmutter, Case No. 1:22-cv-01564-BAH
  • US Copyright Office decision
  • Motion for summary judgment
  • Opposition
• Zarya of the Dawn, Registration No. VAu001480196
  • US Copyright Office decision
  • US Copyright Office Grants Limited Registration for AI-Generated Graphic Novel - Cooley client alert, February 28, 2023
• Does v. GitHub, Microsoft, OpenAI, Case No. 4:22-cv-06823-JST
  • Complaint
  • GitHub/Microsoft’s motion to dismiss
  • OpenAI’s motion to dismiss
• Anderson v. Stability AI, Midjourney, DeviantArt, Case No. 3:23-cv-00201
  
  • Complaint
• Getty Images (US) v. Stability AI
  • Complaint