SoulAuth Wants to Give Your AI Assistant Its Own Permanent ID

A team of seven researchers has released SoulAuth, a free, open-source design for how to keep track of who — or what — is doing things online. Their core argument is simple: as AI shifts from one-off chats to long-running assistants that remember you and act across apps and devices, those assistants need an identity of their own. Not a username. Not your login. Something that survives even when sessions end, devices change, or the software gets copied.

Why this matters to you: AI assistants are already booking appointments, drafting and sending email, and moving money. Right now, almost all of that activity gets logged under your account. If an AI assistant buys the wrong thing, emails the wrong client, or gets tricked by a scammer, the record looks like you did it. SoulAuth proposes separating three things that today get tangled together — who someone is, how they prove it, and what they're allowed to do. Think of it like a driver's license, a car, and a permit to drive a truck: related, but not the same thing.

The paper's unusual move is to treat humans and long-lived AI as equals in the identity system. An AI assistant would get its own ID that isn't replaced when you swap password managers, log in from a new laptop, or restart the server it runs on. That makes it possible to say, months later, which specific assistant took which action. The team built a working prototype in Rust, a programming language known for safety, and released it publicly so others can inspect and test it.

The catch: the authors themselves report only "partial" success. Two big pieces aren't finished — a single unified way to handle passwords and keys, and the ability to trace old actions back to an AI's identity over time. It's also just a research paper, not a product anyone can buy, and no major tech company has signed on. It's an early sketch of plumbing most of us will never see, but will quietly depend on.