Загрузка...

NetClaw, human-in-the-loop approvals #netclaw #openclaw #akkadotnet

People gave AI agents access to their email inboxes, and the agents started initiating legal disputes with their insurers. Not hypothetical, it happened, and we keep a running list of incidents like it on the TextForge blog.

Here is how you prevent it. NetClaw keeps a human in the loop the actor way: a tool approval is a gated message in the mailbox, so nothing irreversible happens until you click approve, even an hour later. Expensive work runs as supervised sub-agents that can't take the whole session down with them. Built on Akka.NET.

Full standup: https://www.youtube.com/watch?v=zdrg987i3vI
NetClaw (source, Apache 2.0): https://netclaw.dev
"OpenClaw gone rogue: five times an AI agent wrecked someone's email" from the TextForge blog: https://textforge.net/blog/openclaw-email-disasters

Видео NetClaw, human-in-the-loop approvals #netclaw #openclaw #akkadotnet канала Petabridge
Яндекс.Метрика
Все заметки Новая заметка Страницу в заметки
Страницу в закладки Мои закладки
На информационно-развлекательном портале SALDA.WS применяются cookie-файлы. Нажимая кнопку Принять, вы подтверждаете свое согласие на их использование.
О CookiesНапомнить позжеПринять