Airbnb's Chesky says AI agents need their own OS
Airbnb CEO Brian Chesky says AI agents need their own operating system — and he's told Sam Altman so. In an interview following Airbnb's fall update, which shipped AI-powered search, Chesky argued that the chatbot is the wrong interface for e-commerce entirely — "what you're seeing today is also not the endgame for e-commerce or for travel or shopping" — and sketched where the company is actually heading: an Airbnb agent in the explore tab, another in customer service, and eventually "a macro Airbnb agent" that will be "interoperable with MCP to other agents." He described agents as "apps with faces" that can interoperate without a business deal first, said chatbots have proven good at lead generation rather than owning the customer relationship, and put the next concrete milestone at multiplayer AI interfaces for group trip planning over the next three to six months, with voice agents rolling out for search and customer service this fall. The sharper claim is the platform one: Chesky says the industry lacks a true OS built for AI, that AI should be operable at the kernel level, and that nobody — including OpenAI — is building the SDK. He told Altman directly that ChatGPT's app store needed "an operating system like the App Store. Otherwise, none of our functionality is going to exist… and of course, it didn't work very well." He's also candid about the current state of agent wrappers: "I've used Airbnb on Muse and Instinct… it doesn't work very well… Airbnbs work even worse. So I don't think we've cracked consumer AI." The take: whoever ships the interoperable agent layer first gets to tax everything above it, and a travel marketplace has an unusually strong incentive to make sure that layer isn't owned by the company that would rather send users to its own answers.
Rep. Pramila Jayapal introduced a National AI Charter Act that would require every AI company in the US to obtain a federal charter before operating. Modeled on bank charters, the framework would ban surveillance pricing, strengthen AI liability, and let the government revoke a charter as a corporate "death penalty" for egregious harms — alongside round-the-clock federal oversight, testing in government facilities, approval before release, adversarial stress tests, and a government-controlled kill switch. "If you want to be an AI company in this country, you have to obtain a public charter that has a broad set of terms and conditions to operate and ensure public benefit without the harms," Jayapal said. Former FTC chair Lina Khan endorsed the framework: "For generations we have required banks, drug makers, and nuclear operators to meet public terms before they do business, and AI companies should be no different." The honest caveat: it's a framework from a progressive backbencher with no stated Republican path, and the White House has opposed AI legislation — so this is agenda-setting rather than near-term law. But it's one of the first attempts to put a concrete licensing mechanism on the table, and it lands the same week OpenAI's own safety leader quit warning that labs aren't careful enough.
Should AI companies be required to hold a federal charter to operate — like banks? Tell us in the comments.
