Let’s break down the acronym. RAG stands for Retrieval-Augmented Generation. In simple terms, it’s a method that allows an AI model to pull from a specific database of information—in this case, the complete discography of AC/DC—before generating new content. Unlike a standard generative model that might produce generic rock sounds, AC/DC RAG retrieves relevant musical segments, chord progressions, and vocal patterns from the band’s library. It then stitches these elements together in novel ways.
The result is music that feels authentic. It has the right tempo. The right key. The right attitude. Some listeners have described it as hearing a lost album from the 1980s. Others find it unsettling, a digital ghost of a living band.
The technology behind AC/DC RAG is not entirely new. Similar systems have been used for text and image generation. But applying it to music, especially to a band with such a distinct and protected sound, pushes the boundaries of what is legally and ethically acceptable.
How the System Works
To understand the controversy, you need to understand the mechanics. The developers behind AC/DC RAG likely trained the model on a massive dataset of AC/DC recordings. This includes studio albums, live performances, and even isolated tracks. The AI learns the statistical relationships between notes, drum hits, and vocal inflections.
When a user prompts the system—perhaps with a description like “a slow blues riff in the style of ‘Back in Black’”—the retrieval component searches its database for the most relevant examples. The generation component then uses these examples as a blueprint to create something new. It’s not copying. It’s mimicking. But the line between inspiration and infringement is razor-thin.
Early demos of AC/DC RAG have circulated on private forums and social media. They showcase tracks that could easily be mistaken for unreleased B-sides. The fidelity is high. The energy is there. But so is the legal gray area.
