End-to-end Encryption

End-to-end encryption works a bit like sealing a letter inside a locked box before sending it. Only the person who wrote the letter and the person meant to receive it hold the key, so even if a courier or the postal service opens the box along the way, they can't read what's inside. No matter how many hands the message passes through, only the original sender and the final recipient can ever actually read it.

What matters here is that this includes the company running the service in the middle. Even if there's a server storing and forwarding the messages, that server doesn't hold the key either, so it can't peek at the content. That's why, whether it's a private conversation or, as has become common recently, multiple AI agents exchanging messages with each other, this method means that even if the conversation is intercepted, the actual content stays locked inside the box.

An article about OpenHuman explains that "messages exchanged between agents are end-to-end encrypted using the Signal protocol." A common misunderstanding here is that end-to-end encryption doesn't mean a conversation is completely safe from hacking or errors. What this technology prevents is strictly limited to someone in the middle viewing the content — whether the other party in the conversation (in this case, each person's clone or agent) can itself be trusted is a separate issue.