From the founder

I built Akaeon with the goal of protecting digital authorship. Your work can travel anywhere in the world instantly, and locking it behind a paywall denies you the chance at discovery while denying everyone else the chance to enjoy it. The price of control is now reach, but it’s not a price we have to pay. People deserve credit for their work regardless of reach.

“Our content may not be used for AI training” is no longer sufficient. A line in a terms-of-service can be ignored by a crawler, or waved away afterwards by arguing the clause was added once the training run had already finished. robots.txt, ai.txt, TDMRep files and Content Signals are all silently mutable by anyone with write access. Two years into a dispute the publisher cannot prove what the file said on a given date, and the lab cannot prove what it saw. Both sides end up arguing from memory about a document that has no history.

The theme over the past decade has been to make our lives more convenient. Google Maps, Uber, one-day Amazon delivery, and now AI. The question becomes how much of our cognitive capability we are willing to outsource in the name of convenience. I firmly believe authorship and creativity should not be on that list.

So we developed a solution by combining Merkle trees, DNS challenges and immutable timestamps into a single product that works for a publisher with a proprietary licensed dataset and for an individual with one piece of art. So authorship can now be proved, not just claimed. Permissions are not black and white, so our records carry separate flags for separate uses. With one file you can reach as many people as you wish while still declining reuse or AI training. With any record you need its history, so nothing gets deleted, only updated. You always know what you allowed and when. All of this means we built a registry so transparent and simple that even your grandmother can find a record.

All of our work was built on top of a research foundation. Independent scientific studies measuring whether commercial image-to-image models leave recoverable behavioral fingerprints under near-imperceptible adversarial perturbation. What pressures lead foundation models to cave to sycophancy. At Akaeon our goal is to work out what can actually be verified and what we are only asserting. So we can build a more transparent web.