The Independent Music Publishers International Forum, more succinctly known as IMPF, has put forward a proposed set of guidelines for enforcing ethical practices in the fast-expanding realm of AI technology. Here’s a breakdown of the group’s plan, signed off on by over 200 members, including Concord Music Publishing, Big Machine Music, Downtown Music Publishing, Reservoir and more.

Key components: The IMPF’s proposition rests on four principles, each designed to maintain transparency and copyright equity among songwriters, publishers, and developers of AI music models. Take a look:

• Developers must seek express permission for the use of music in the machine training process, without exception – this will protect creators’ economic interests as well as their right to expression.

• Comprehensive records of all licensed musical works used in the machine training process must be documented, ensuring that any remuneration be accurately distributed to all credited writers and rights holders.

• Clear labels differentiating work created by humans vs. work generated by AI should be applied to maintain a “level playing field for human-created music while protecting consumer choice.”

• Developers should also transparently differentiate between assistive AI and fully generative AI applications.

Why it’s important: The motion comes toward the end of a year marked by both major innovations and industry panic within the AI space, evidenced by the emergence of fake songs passing as original creations by A-list artists as well as ensuing copyright litigation. The IMPF’s push for guidelines marks one of the biggest steps forward in balancing artist rights with technological advancement.

What they’re saying: “[AI] will only get more sophisticated,” stated IMPF President Annette Barrett. “In many cases, it will actually enhance our work and lives. We should not fight these advancements, but it would be negligent to give tech developers free reign when it comes to the use of artistic human work — which carries its own irrefutable, intrinsic value — to enable machine learning.”