Meta Muse agent was in over half of the sites that Cequence studied
Cequence tells that Muse called itself a usual Chrome browser, so a site cannot tell it from a person.
Cequence Security tells that the Meta Muse AI agent was in the traffic of more than half of the customer sites it studied. The study took about three weeks. At the median customer, Muse traffic went up almost six times in about two weeks. Cequence also made Agent Trust, a tool that finds AI agents and controls single steps.
What Cequence saw
Muse did not tell that it was an AI agent. It used a Chrome browser in the cloud and a consumer VPN (a private network). Cequence tells that Muse signed in to bank accounts and passed multi-factor checks for users. It also filled baskets, and a small part of the sessions ended in a purchase.
How Agent Trust works
The tool keeps a live list of agents and checks their identity. It can block, slow, or test one message and let the agent continue. For large changes to an account, it can tell the account holder to make sure on their own device.
What is not known
Cequence gave no names of customers. One travel firm blocked almost one in five Muse messages. Cequence did not tell how many sites it studied.
Insight: Muse brings the sales and the problem
Earlier reports tell that Muse has more than 10 million users and that Meta plans to get a part of the sales made with it. Muse also looks like a Chrome browser. It looks like Meta gets the sales while sites carry the cost to find the agent, and Cequence has the tool for that cost. Sites must set agent rules.
Sources
- Meta's Muse AI agent seen in half of customer trafficSecurityBrief UK
Posted