Meta, Walmart, Stripe and several other companies are working on a new standard for AI agents that could make it easier for these tools to interact with businesses. This project will be known as “personal agent protocol,” which aims to create common rules for how AI agents identify themselves, access websites and handle customer information.
The initiative is being led by Bret Taylor, co-founder of AI company Sierra and chairman of OpenAI. He said businesses need a clear way to know whether they are dealing with a real customer or an AI agent acting for that person. The need for such rules is growing quickly. Meta’s Muse, launched in September, has become a popular AI agent, with millions of users in the US. People use it for everyday tasks such as booking appointments, finding gifts, and signing children up for classes.
But AI agents have also created concerns for businesses. Amazon has blocked Meta’s agents and has also taken legal action against Perplexity over concerns about AI tools accessing its website without permission. Security is another major issue. Meta executive David Singleton said AI agents may need access to personal details and payment information, so users should have more control over what these tools can do.
The companies hope the new standard will work like email, a common system that lets different services communicate. OpenAI and Anthropic have not joined yet, but Taylor expects more major AI companies to take part. The goal is simple: make AI agents work with businesses in a safer, smoother and more organised way.
