Join the #ImagineBetterAI conversation on our social media channels. You can also sign up to receive our monthly AI Observer newsletter here.
Welcome back to the AI Observer, HLK’s newsletter tracking the AI legal landscape.
In the latest instalment, our experts explore the ongoing legal battle between major record labels and AI companies Suno and Udio over the use of copyrighted music in generative AI models.
Our experts also report on the recent legislative developments in AI including the opposition to a California bill for regulating powerful AI models, OpenAI’s text watermarking tool and a YouTube content creator who has filed complaints against OpenAI, Google and Nvidia.
Did you miss the previous issue of the AI Observer? Check out all our previous issues and other AI updates here.
AI faces the music
In June this year, a new front opened up on the copyright war against current generative AI practices, and this one is set to music.
In this article, Caroline Day discusses the ongoing legal battle between major record labels and AI companies Suno and Udio over the use of copyrighted music in generative AI models.
Read the full article here.
Legislation watch
A California bill for regulating powerful AI models faces opposition as it enters the final legislative phase. If passed, the bill would mandate that where an AI model is trained with more than $100 million worth of computing resources the developers must disclose and follow a safety and security protocol to prevent the AI model from causing “critical harm”, for example damages exceeding $500 million. The bill would also provide protections for whistleblowers in AI companies who speak out about the risks in the AI models being developed. With California home to most of the top AI companies, including OpenAI, this bill is one to watch, and it could inspire and strengthen AI regulations elsewhere, including the EU. The European Commission’s AI Office has been reported to have been closely following the bill and to have met with Senate representatives in the summer to exchange views and explain the EU AI act.
OpenAI recently confirmed that it has developed a text watermarking tool that would identify AI-generated content, which could be implemented in Chat-GPT. The watermarking tool, rather than leaving a visual trace, would leave a digital stamp enabling its accurate detection as machine-generated content. This tool would help towards compliance with the EU AI Act (the specific transparency risk clauses), but OpenAI has expressed doubts circumvention of the system, for example by using another LLM to reword the content. There is also the understandable worry that users would favour other LLMs instead of Chat-GPT to avoid the watermarking.
Three related Complaints filed in the US against OpenAI, Google, and Nvidia by a YouTube content creator are worth noting because none of them includes claims of copyright infringement. Instead, the claims are for unjust enrichment and unfair competition in relation to the companies using transcripts of millions of YouTube videos to train their machine learning models. These cases represent the continued questions around the way that training data for machine learning models is scraped from the internet.
AI application of the month
Meet BirdNET, your pocket-sized birdwatching companion. This clever AI app identifies birds just by their songs. Hit record, and BirdNET’s neural network swoops in, analysing the chirps and tweets to tell you which bird species are currently serenading you. Developed by the Cornell Lab of Ornithology and Chemnitz University of Technology, BirdNET not only recognises the usual garden suspects but also considers your location and the time of year to inform the identification.
Whether you’re a seasoned birdwatcher or just curious about what’s in your garden, BirdNET makes bird identification simple while helping scientists monitor avian populations for conservation efforts.
This is for general information only and does not constitute legal advice. Should you require advice on this or any other topic then please contact hlk@hlk-ip.com or your usual HLK advisor.
