Toby Gad: Keep AI as a tool, not a replacement for songwriters
Toby Gad doesn't think AI-written songs help the music industry. The Grammy-nominated songwriter and producer behind hits for Madonna, John Legend, and Beyoncé draws a hard line: use AI to support creativity, not to outdo humans.
His stance is simple. There's helpful AI that makes the process sharper - idea generation, organization, faster iteration. Then there's AI that tries to "write songs better than us." He's not on board with that.
The fault line: assistant vs. substitute
As Gad puts it, "There is the type of AI that will make the computer smarter and help us to be more creative. That's the AI I love." The problem starts when AI aims to replace the artist: "I don't like computer-written songs and computer-generated art that competes with ours. I think we should preserve the human artistic expression."
He's also worried about where this leads. If AI undercuts human authorship at scale, it's bad for culture, not just careers.
Artists are pushing back
Radiohead's Thom Yorke argues today's systems remix human work without true originality or proper credit. "It analyses and steals and builds iterations without acknowledging the original human work it analysed. It creates pallid facsimiles… the economic structure is morally wrong … writers are not paid." He called it a "weird kind of w****, tech-bro nightmare future."
He was one of 10,500 signatories - including ABBA's Björn Ulvaeus, Julianne Moore, The Cure's Robert Smith, and Rosario Dawson - warning AI companies that unlicensed use of creative work is a "major, unjust threat" to artists' livelihoods.
What this means for working creatives
- Use AI as an assistant, not a ghostwriter. Idea prompts, references, arrangement experiments, admin - sure. The melody, lyrics, performance, and taste stay yours.
- Protect your process. Keep drafts, stems, timestamps, and credits. If you collaborate, write down who did what.
- Watch what you upload. Don't feed unreleased or rights-controlled work into public tools. Assume anything you paste could be stored or learned from.
- Pick rights-aware tools. Favor vendors with clear licensing, opt-outs, and attribution. Review the U.S. Copyright Office guidance on AI to stay compliant.
- Lock it into contracts. Add clauses that forbid partners from training AI on your work without consent and payment. Require warranties that no unlicensed training was used.
- Make your fingerprint obvious. Story, phrasing, micro-imperfections, and recurring motifs. AI imitates; you differentiate.
- Upskill on ethical workflows. If you're integrating AI, learn smart, defensible practices. See our AI courses by job for structured options.
- Join collective action. Guilds, unions, and cross-industry initiatives can set standards for credit and compensation.
The takeaway
AI that amplifies your craft is useful. AI that replaces authorship undercuts the point of art. Use the first. Push back on the second. Keep the human voice at the center - that's the value no model can fake for long.
Your membership also unlocks:
