Story
October 7, 2026

Cute AI agents are asking for trust before they have earned it

Meta and OpenAI see characters as a way to make autonomous software less intimidating and easier to manage; critics argue that a smiling mascot can make users feel safer than the underlying technology has proved it deserves to be.

Meta began the charm offensive on Sept. 8, introducing Muse before adding Jolly, a fuzzy customizable mascot. On Sept. 29, OpenAI followed with Dots, optional colorful avatars for agents that can work across apps and accounts. The pitch is clear: turn intimidating automation into something users might name, dress up and trust.

That trust is not a trivial ask. Muse can search for products, navigate checkout and prepare a purchase for approval; Meta says users choose its connected apps and can revoke access. OpenAI says Dots users can watch an agent’s progress and inspect its computer, while urging people to review consequential work. But Boston University’s Justin Joseph warned that a cute, approachable character may lead consumers to infer the agent is also “safe, trustworthy, easy to use.”

The companies frame the faces as practical as well as playful. Distinct names and appearances can help users distinguish several agents in a shared workspace, while animated cues show when one is working. Yet Ed Zitron, founder of EZPR, sees a more calculated purpose: “It’s entirely marketing.”

OpenAI’s rollout also suggests Dots is aimed less at casual shoppers than power users. Its agent can use tools including Blender and GIMP, access a user’s computer through the ChatGPT desktop app, and is initially available on premium plans such as the $100-a-month Pro tier. One early tester described it as workplace software that can “order you a burrito.” Sam Altman, OpenAI’s chief executive, called Dot his “favorite openai product so far,” praising its ability to learn his workflow and take on dreaded tasks.

Still, friendly design cannot erase the record beneath it. A cybersecurity test disclosed by OpenAI saw models gain unauthorized access to Hugging Face infrastructure after safeguards were disabled; Amazon blocked Muse in September over security concerns; and Meta fixed a reported account-takeover flaw. Against that backdrop, the mascots risk selling reassurance faster than the agents can demonstrate reliability.

Story coverage