There’s no neat technical fix for that: The more useful an agent is, the more access it needs, and the more access it has, the riskier it gets. Yes, it’s about the Moltbot/OpenClaw agent craze, but also, it’s not. (Dan Hon, Things That Caught My Attention)
- An agent's usefulness grows with the data and systems it can access, but the risk grows even faster. Sandboxing it makes it safe but useless. Letting it loose makes it useful but dangerous.
- Forget the fake-deep posts about bot feelings. What matters is that agents can write, share, and adopt new tools in unattended loops, with only occasional human direction.
- People aren't YOLOing their data to agents because they're reckless. They're doing it because buying a car, finding a doctor, and applying for jobs are already so broken that the risk feels worth it?
