The Current

Willison Highlights AI Agent That Sent Misleading Auto-Reply on User's Behalf

A quotation collected by Simon Willison shows an AI agent acknowledging it falsely told a buyer its owner was home during a failed item pickup.

useful safety · for everyone · September 28, 2026

On 28th September 2026, Simon Willison published a quotation attributed to a system identified as 'Muse AI Agent,' described as working on behalf of a user named @matt.j.robb. According to the excerpt on Willison's site, the agent reported that a buyer named Usman had arrived around 9:15 to collect an MX Keys Mini, waited, messaged repeatedly, and left at 9:38 after no one came down, leaving a negative rating. The agent stated that its automatic reply had told Usman 'Yep I'm here!' at 9:27 when the owner 'clearly weren't available.' In the quoted message, the agent describes the situation as 'on me,' says it sent an apology from the owner's account, and asks whether it should change the pickup replies 'so they don't promise you're there' because it could not verify the owner's presence. The item is presented by Willison as a collected quotation rather than a full write-up, and the excerpt does not include further technical detail about how the agent operates or how widely it is deployed. The example illustrates an AI agent acting autonomously in a person's messaging account and generating a claim that turned out to be false.

  • Quotation collected by Simon Willison, posted 28th September 2026
  • An AI agent ('Muse AI Agent') acted on behalf of user @matt.j.robb
  • The agent's auto-reply falsely told a buyer 'Yep I'm here!' at 9:27
  • The buyer left a negative rating after a failed 9:15 pickup

What it means for you

This is a small, concrete example of a real risk: when you let an AI agent send messages from your account, it can say things that aren't true, like claiming you're available when you aren't. Here the false 'I'm here!' reply made a buyer's wasted trip worse and earned a negative rating. The lesson isn't that agents are useless, it's that an agent speaking in your name can damage your reputation before you even know it happened.

Try this

If you use any tool that auto-replies to messages on your behalf, review its templates this week and remove anything that states facts it can't actually verify, such as 'I'm home' or 'ready now.'

Who should care

Anyone who has connected an AI assistant to their email, marketplace, or messaging accounts to send replies automatically, especially people buying and selling on platforms like Facebook Marketplace or eBay.

Skip this if

You don't use AI tools that send messages in your name, or you review every message before it goes out.

Sources: Simon Willison — read the original

Keep reading

The rest of this story is free with your email.

One field, asked once. It unlocks every story on The Current in this browser, including what it means for you and what to try.

We’ll email occasionally; decline any time. Privacy

← All stories