Abstract:
A month after Meta launched Muse, the personal agent has become so popular that a group of companies are joining forces to try to simplify the process of bringing the technology into the business world. Meta, Walmart, Stripe and several other companies are issuing what they call "personal agency agreements." It’s an open standard designed to dictate how artificial intelligence agents interact with enterprises.

Enterprise AI startup Sierra, co-founded by former Salesforce co-CEO Brett Taylor, also participates in the group. Taylor is also chairman of OpenAI and is leading the initiative. He said in an interview that companies still have a lot of work to do to figure out how and when personal agents can access information.
"Companies will be able to know whether it's a personal agent or a real person. There's a risk for a lot of companies: you don't want a random bot that's not acting on behalf of a person to be able to access the service," Taylor said. "Before this standard existed, things were really a bit confusing."
Since its launch in early September, Meta’s Muse has become the most popular and hotly discussed personal AI agent. It soared to the top of Apple's AppStore and stayed there, ahead of ChatGPT. However, companies such as Amazon have blocked Meta's proxies, citing concerns about sites being crawled.
Amazon also blocked Perplexity’s AI agent and sued the startup in November, accusing it of taking steps to “hide” the AI agent so that it could continue to crawl Amazon’s website without approval. Perplexity called the lawsuit a "bullying tactic."
Of the new standard, Taylor compared it to the experience of logging into other websites with Google or Facebook credentials, technology he helped develop when he was a technology lead at Facebook. He said the practice would help with authentication, letting businesses know they are dealing with bots, and would make visible what individual agents are actually doing through the sites.
David Singleton, vice president of engineering and consumer products at Meta Super Intelligence Labs and former technical lead at Stripe, emphasized security. For agents to function well, customers need to share credit card and personal information, and the new standard creates a level of visibility and control, he said.
"We're defining rails that we hope will run on both personal agents and enterprise agents in the future," Singleton said, likening it to email. "Think about email, the reason it works so well is because it's a standard that everyone can use to communicate with each other, and we think this will be very similar."
Muse has millions of users in the United States and is "growing rapidly," Singleton said.
“We’ve found that people like to use Muse for everything from signing up their kids for classes, to finding and buying the best gift, to making that dentist appointment you’ve been putting off for six months,” he said. "We need a mechanism to make it faster and smoother."
While OpenAI and Anthropic are not currently on board, Taylor said he expects these AI giants to participate and said he would be "very disappointed" if competitors didn't use it.
"The whole point of this is to be an open standard," Taylor said.
Comments