Listen in the Podli app 🎧
Follow your favourite podcasts, listen offline and in the car with CarPlay and Android Auto, and always pick up where you left off. Free to try.
In this edition, Host & audio engineer Ashea is joined by Production Expert Founders Mike Thornton and Russ Hughes & post-production audio engineer Paul Maunder, to address one of the most significant regulatory changes affecting the audio and music production industry: the new EU AI Act. With implementation now in force, this episode cuts through the hype and confusion to explain exactly what the law requires, how it applies to audio professionals, and what both individuals and platforms must do to comply.
The conversation covers the practical implications for anyone using AI voice generation (like 11 Labs), AI-assisted editing tools, or working with platforms like Spotify and YouTube. The team breaks down machine-readable marking requirements, disclosure obligations, audio watermarking technology, platform accountability, and what the experience with GDPR implementation can teach us about enforcement. This episode is essential for every audio professional seeking clarity on regulatory compliance and industry integrity in the AI era.
In This Episode:
- EU AI Act Overview: GDPR for AI — How the EU AI Act functions as 'GDPR for artificial intelligence,' applying to any content played in the European Union regardless of where it was created
- Deepfakes Defined: Beyond Celebrity Videos — The law's broad definition of 'deepfakes' as anything that could be misconceived as human-made but is AI-generated, including voice synthesis, video, and images
- Machine-Readable Marking Requirements — How AI system providers must embed technical metadata into AI-generated content so it can be automatically detected by machines and platforms
- Disclosure Requirements for Deployers — Content creators and distributors must explicitly disclose to audiences when content contains AI-generated or manipulated audio/video
- The Disclosure Chain: Creators to Audience — How disclosure cascades through the value chain: if you use 11 Labs for voiceover, you tell your client, who tells their audience
- AI Cleanup Tools Exemption — The law doesn't apply to AI-assisted editing or enhancement tools unless the regenerated content becomes the primary focus
- Pro Tools Generative Features Impact — Real-world example: if you use generative AI to regenerate missing words in a voiceover dropout, you must disclose this
- Spotify's Rapid Implementation — Within days of the EU Act coming into force, Spotify implemented labeling showing listeners when content uses AI
- YouTube's Platform Immunity Defense — YouTube's claim that they're merely a platform; how this misinterprets the law and shifts burden inappropriately
- Platform Responsibility vs. Creator Responsibility — Why YouTube, Facebook, and Instagram should be held accountable as publishers of content, not just neutral platforms
- Claude's Watermarking Response — How Claude added digital watermarking to AI text within days, demonstrating rapid compliance by major AI providers
- Machine-Readable vs. Human-Perceptible Disclosure — The distinction between technical watermarking (for machine detection) and human-friendly disclosure (text/audio announcements)
- Audio Watermarking Technology: Silent Cypher — Sony's deep audio watermarking technology that embeds imperceptible marks in audio frequency ranges
- Audio Brain for Mac: Emerging Audio Watermarking — New tools providing audio watermarking specifically for Mac OS, representing rapid tool development for compliance
- Complex Mixes and Watermarking Challenges — How tracking individual AI-generated elements within massive film mixes presents technical detection challenges
- GDPR as Enforcement Precedent — Since 2018, GDPR fines total ~€7 billion; this historical data proves regulatory enforcement intention
- Big Players vs. Small Businesses — Enforcement targets Spotify, YouTube, Facebook—not individual creators; the law aims to bring major platforms into compliance
- Copyright Protection Gaps — The EU Act and US No Fakes Act don't adequately address copyright of training data and content protection
- US Copyright Office Stance: Prompt-Driven AI — US Copyright Office denies copyright to purely prompt-driven AI output; Invoke AI's successful appeal shows lines are still being drawn
- Creative Use of AI vs. Deceptive Use — Distinguishing between legitimate transparent creative use (like T-Pain's autotune) vs. deceptive undisclosed use
- Professional Responsibility: Setting Standards — How professionals in audio should embrace the law and set standards themselves rather than waiting for enforcement
- Why Imperfect Laws Still Matter — No law is perfect, but laws represent our best effort at improvement and societal standards
- Race to the Bottom Without Standards — How lack of regulatory enforcement creates industry pressure to adopt AI, lower rates, and compromise quality
- Hypocrisy in Industry: The MPSE Poster Incident — Example of major audio professionals organization using generative AI while advocating against AI job displacement
- YouTube's Selective Compliance — How YouTube selectively enforces responsibility when reported violations occur, claiming external responsibility
About Our Guests:
Mike Thornton:
Co-founder of Production Expert and host of the Production Expert Podcast. Mike serves as industry commentator and facilitator of important conversations between technology companies and the professional audio community. In this episode, Mike moderates discussions on AI regulation and compliance.
Russ Hughes:
Co-founder of Production E
