6 chapters across the catalog
A reported hack of Suno, an AI music generator, revealed the company used Podcast Index to target content for its training models. The generator identified 420,000 podcasts with at least five 30-minute episodes, attempting to download approximately one million hours of audio. This automated bot activity resulted in a significant bandwidth spike for the Podcast Index public assets page, consuming roughly five terabytes of data.
Dave Jones expresses concern about the data traffic burden on small apps. Adam Curry jokes about a "two-stage rug pull" involving lifetime subscriptions that later turn into monthly fees, while the "Pod Sage" (Dave) considers how to build a workflow that doesn't overwhelm app developers.
John Spurlock argues against discriminating based on the caller, noting that "one man's bot is another man's app." He provides data showing that Buzzsprout and Red Circle are the most aggressive at blocking media requests, while most other hosting companies remain open.
Dave Jones argues that hosting companies have a contractual obligation to creators, not to abusive bots that rack up terabytes of bandwidth costs. He cites the scale of HLS abuse as a significant threat to independent hosts like Pinecast who may not be able to afford the resulting bills.
John Spurlock argues that much of the current industry anxiety is based on the fear of potential HLS abuse rather than documented reality. He encourages the community to look at actual data before making restrictive decisions that could compromise the open nature of podcasting.
Hosting companies like Buzzsprout are concerned about the financial impact of bandwidth abuse from bots downloading HLS video streams. A proposed solution involves using HTTP signatures and public keys, similar to the Mastodon ActivityPub protocol, to authenticate legitimate human users and apps before allowing video downloads.
