A Hacker News user bought sheet music by Erik Satie for a PianoDisc Protigy, a self-playing piano, then ran Astra and Fable AI systems against each other, feeding one's output into the other as critique. The resulting question, posted September 5, is blunt: can the results be released publicly. No one in the thread offers a citation to a license, a terms-of-service clause, or a court ruling. That gap is the story.
The person did not break any law to get here. Satie's compositions are long out of copyright, and running software on a piano someone owns is not, on its face, infringing. What is unresolved is the layer in between, the AI-generated critique and whatever transformed audio or text came out of the back-and-forth between Astra and Fable. Neither company's terms are quoted in the thread, and the poster does not say which service, if either, claims ownership over outputs derived from user-supplied audio.
For product and design leads, the stakes are not abstract. A support ticket like this is a preview of what happens when two separate AI products are chained by a user in ways neither vendor tested or authorized. Astra and Fable were not built to interoperate, and nothing in the post suggests either company anticipated its output being piped into the other's input. That means any liability language in their respective terms was written for a single-product use case, not a daisy chain, and may not cleanly apply.
Check what the terms actually say about derivative outputs, not what marketing implies. A clause that assigns output ownership to the user typically assumes the user generated that output through the vendor's intended flow, inside the vendor's own interface. Once output from one system becomes input to a second, unaffiliated system, the chain of custody for rights gets murky fast, and most terms of service do not contemplate it. This is not a hypothetical edge case. It is a one-line prompt away from happening inside any product that accepts arbitrary text or audio input.
Nobody answered the Hacker News post with a definitive legal reading. The closest thing to resolution is the fact that the question was asked at all, on a forum read by the people who write these terms. A self-playing piano, a public-domain composer, and two AI tools arguing with each other produced something nobody has a clean answer for yet. Until a vendor updates its terms to address chained AI outputs explicitly, the answer to the user's question stays exactly where it started: unresolved, in a comment thread.