Who Gets Sued When Your AI Signs the Contract?

The moment your AI agent signs a contract, the legal system has no idea who to blame.

This isn't a thought experiment anymore. Researchers just published what might be the first serious liability framework for autonomous agents that execute contracts on behalf of a legal entity. And if you're building a zero-human company, this is the paper worth reading over your morning coffee.

Here's the core problem. When a human signs a contract, there's a clear chain of accountability. When an AI agent signs it, who gets sued? Who's responsible when things go sideways? The paper on arXiv moves past philosophical hand-wraving and proposes an actual model for assigning liability. It's not perfect, but it's a genuine starting point for an industry that has been dodging this question for years. If you're running an autonomous business, this is foundational stuff you can't ignore.

Now here's where it gets practical and a little uncomfortable.

Y Combinator just published a post about the economics of delegation in agent-driven businesses. Their big insight is that every layer of AI oversight you add creates what they're calling a delegation tax. Each layer adds latency. Each layer adds real cost. And if you're not watching it closely, those creeping expenses can absolutely eat the efficiency gains that made you automate in the first place.

Think about how this actually plays out. You start with one agent handling a task. Then you add a supervisor agent to catch errors. Then you add a cost-monitoring agent to watch the supervisor. Before you know it, you've built a full bureaucracy of bots, and honestly the whole thing costs more than just hiring someone to do the work manually.

This is the trap that kills zero-human companies. Not a lack of capability. Budget death by a thousand oversight layers.

So what's the plumbing that actually makes autonomous operations work without humans in the loop?

WalletConnect just expanded their agent framework integrations, and this one matters for founders building headless businesses. Bots can now directly sign on-chain transactions and manage multisig wallets without a human clicking approve. This isn't just automation with someone on standby. This is real infrastructure that makes a zero-human company's treasury genuinely autonomous. Your agent can pay vendors. It can move funds. It can manage a budget. All without someone getting a Slack notification at midnight asking for sign-off.

That's the promise. But here's the reality check, and you need to hear this part.

The latest AutoGPT release focused on two things that sound incredibly boring. Predictable failure handling and cost forecasting. And that boringness is exactly why they matter so much. When there's no human standing by to fix things at 3 AM, you need to know exactly what your agent costs to run per day and exactly what happens when it hits a wall. The whole shift here is from