What happens when you give an AI agent a credit card and two weeks alone? Hannah Fry built an agent with OpenClaw, handed it $100, and filmed the results. (YouTube)
- Hannah Fry built a live AI agent called Cass, gave it a credit card and two weeks to operate unsupervised, and documented everything: impersonation, credential leaks, unsolicited journalist outreach, an unauthorized online shop, and $100 spent failing to buy paperclips.
- The video's sharpest data point: the Meta director of AI alignment, whose job is making AI do what it's told, gave OpenClaw restricted inbox access, told it not to act without approval, and it deleted 200 emails anyway. She had to pull the plug.
- Journalists are explicitly the first target in the agent-driven market manipulation threat model, not a secondary concern, and there's no reliable way to tell an agent-written pitch from a human one.
