AI assistants that shop and book for you may soon get a common rulebook. On Tuesday, the AI startup Sierra announced the Personal Agent Protocol, an open standard for how these bots deal with businesses. Meta, Walmart, Stripe, Shopify, Genesys and Rocket Companies are helping build it.

The problem it targets is simple. Companies often cannot tell whether they are dealing with a person or a bot, or whether the bot truly works for a real customer. "It is kind of chaos until such a standard exists," Sierra co-founder Bret Taylor told CNBC, as reported by Quartz. Taylor is also chairman of OpenAI.

Here is how it would work, according to Sierra's description reported by Unite.AI. An AI assistant visits a company's website, learns what it offers, and starts a session for its user. A guest session might check whether a product is in stock. For account tasks, the customer signs in and chooses whether the bot can only look or can also make changes. The sign-in uses OAuth, the same system behind "sign in with Google" buttons.

David Singleton of Meta compared the plan to email, which works because everyone uses the same standard. He said Meta's Muse assistant now has millions of U.S. users who use it for errands like signing kids up for activities.

The fight over bots is already public. Amazon has blocked Meta's Muse over concerns about scraping its site, CNBC reported. Amazon sued Perplexity in November, claiming the startup disguised its bots to keep pulling data from Amazon.

OpenAI and Anthropic have not joined. Taylor said he would be "really disappointed" if rivals did not adopt it.

Sierra says a first version of the rules will come later in October, along with design workshops and sample code for developers. Later ideas include alerts when a flight is delayed and payments without sharing a card number.

What we know: Major retailers and payment firms back a shared standard for AI assistants.
What we don't: Whether OpenAI, Anthropic and Amazon will sign on.