Are we too comfortable giving AI agents access to our personal accounts?
TheObeisantLaurels · reddit · 2026-10-12
A Reddit user questions how much personal access AI agents need: email, calendars, browser sessions, contacts, even payment info. While testing an agent that fills forms and signs up for services, they realized it was submitting their real email and phone to random sites. The thread asks how builders handle privacy—isolated accounts, restricted permissions, sandboxes, or plain trust.
More from coding & agent
- Developer trains Codex to punt every hard task to Opus 5.5 — and it works — jdjohnson · 2026-10-12
- Defining architectural shape is the key to spec-driven AI coding — gregmushen · 2026-10-12
- UI Tip: Group Background Job Toasts Into One — but Never Hide Errors — jonathan_wilke · 2026-10-12
- Not maker mode, not manager mode — the secret third thing: juggling agents — nbaschez · 2026-10-12
- Rigging in Blender via MCP: six-step recipe for Opus 5.5 to build and animate 3D scenes — sidahuj · 2026-10-12
- Agentic hacking looks like breadth-first search over misconfigurations, not exploits — dyn___ · 2026-10-12