OpenClaw's creator just published his monthly AI bill: $1.3 million. OpenAI i...

OpenClaw's creator just published his monthly AI bill: $1.3 million. OpenAI is footing the entire thing. He's running 100 AI agents with a team of three people, burning through 603 billion tokens in 30 days. His take? "I can disable fast mode and it's 70% cheaper. So it's more like one employee."

That's not a flex. That's a guy already planning for the price correction.

Anthropic didn't wait. They locked down their SDK and raised prices this month. The cheap-AI window is closing. Every workflow you've built on subsidized pricing is going to get a real invoice.

Here's the part that should worry you more than the price tag: that same project is sitting on 512 acknowledged security vulnerabilities. This week, researchers chained four of those flaws into a single attack path (data theft, account takeover, sandbox escape, persistence). One chain. Four flaws. Active exploit advisory circulating right now.

So the actual cost of running AI agents isn't just tokens. It's tokens plus every vulnerability you created by handing software tools nobody audited.

Business owners adopting agent tooling are absorbing two risks they haven't priced in: the subsidy disappears, and the attack surface they've created doesn't.

If you're running AI agents in your business (or about to), here's the question worth asking: what does your workflow actually cost when the subsidy disappears and the first vulnerability gets exploited?