← Accueil actu | Salle du marché
2026-09-14 · Source:TechRadar

A hands-on trial of Meta's new Muse assistant found it remarkably effective at digital errands, though its usefulness hinges on granting broad access to personal information.
Meta's new Muse AI agent has drawn a strikingly split first impression from TechRadar, whose hands-on review calls the assistant incredibly useful at running digital errands while also describing the act of surrendering one's digital life to it as deeply uncomfortable.
The review's central finding is a stark trade-off: Muse only proves its worth once a user hands over a volume of personal information the tester describes as unsettling. The agent impressed its tester precisely because it could act on the user's behalf across everyday digital tasks, yet that same capability required entrusting the system with a sweeping view of private data. The feature and the risk are the same mechanism — the more an agent can see and do, the more valuable it becomes, and the more invasive the arrangement feels.
That tension is the defining challenge for agentic AI products as they move from demos to daily use. A hands-on account like TechRadar's matters because it tests not just whether the software works, but whether a reasonable user would accept the terms. The reviewer's verdict — incredibly useful, deeply uncomfortable — suggests Muse clears the capability bar with room to spare while leaving the trust question wide open.
For the broader market, the episode underscores that the sticking point for personal AI agents may be less about raw performance than about permissions. Utility in this category is purchased with access, and access is precisely what the reviewer found hardest to hand over. Whether Meta can make that exchange feel safe, rather than merely powerful, will likely shape how far Muse travels beyond its earliest testers.
Mots-clés: Meta · Muse AI Agent · AI Agents · Privacy
萬安算交所 AIXX · AIXX · Bourse compute vérifiée · Accéder à la salle