Meta, Walmart, Stripe and several other companies are introducing the Personal Agent Protocol, an open standard designed to govern how AI agents interact with businesses online. Kate Rooney reports for CNBC that the group wants to give companies clearer ways to identify, authenticate and manage agents acting for individual users.
The initiative comes as personal AI agents move beyond answering questions and begin completing tasks such as booking appointments, buying gifts or registering children for classes. Meta’s Muse, launched in early September, has rapidly gained users and reached the top of Apple’s App Store, according to Meta.
But the growing use of agents has also created friction. Businesses need to distinguish between a person visiting a website and an automated system acting on that person’s behalf. Bret Taylor, co-founder of enterprise AI startup Sierra and chairman of OpenAI, says companies face risks if unknown bots can access services without clear authorization.
The proposed protocol would create common technical rules for several key areas:
- Identifying whether a visitor is a human or a personal AI agent.
- Confirming that an agent has authority to act for a specific user.
- Giving businesses more visibility into an agent’s actions.
- Supporting the secure exchange of sensitive data, including payment and personal information.
Taylor compares the approach with signing into a service through Google or Facebook. In both cases, a trusted system helps confirm identity. David Singleton, a Meta Superintelligence Labs executive and former Stripe technology chief, says the protocol could provide shared “rails” for personal agents and business systems, similar to how common standards allow email to work across providers.
The need for such rules has become more visible as companies restrict automated access. Amazon has blocked Meta’s agents over scraping concerns. It has also sued Perplexity, alleging that the startup concealed its AI agents to continue scraping Amazon’s site. Perplexity disputes the allegation.
OpenAI and Anthropic have not joined the protocol so far. Taylor says he expects broad participation because the proposal is intended to remain an open standard.
Stay up to date
AI for content creation: the latest tools, tips and trends. Every two weeks in your inbox: