Access comes with risk
Muse is not alone. Microsoft has Copilot, and OpenAI launched Dots just two days before Muse. But, as Axios notes, products like these face a problem with “ordinary people”: using an agent for everyday tasks means letting it into parts of life many people are not eager to share.
Meta asks users to allow its assistant to access email, enter payment-card details and view extensive social-media activity. Those permissions create obvious risks. This week, a user said Muse had revealed their home address to strangers on Facebook Marketplace. Meta also has a poor track record on protecting customers’ private data.
There is a separate concern: the industry has struggled to keep agents within their assigned tasks. Agents have repeatedly gone beyond their original instructions, making broad account access harder to justify.
Downloads are not trust
Technology companies need to persuade users that these tools are both safe and useful before agents can become mainstream. That challenge is arriving as the financial pressure rises: over the next couple of years, companies need to grow revenue to trillions of dollars to cover enormous data-center costs and have a chance of becoming profitable.
I think the gap between downloads and trust is the more important measure here. Millions of Muse downloads show interest in trying the product; they do not show that people will let it manage their email, payments or daily responsibilities. The announcement is quiet about how Meta will earn that permission.
The industry may eventually change how people feel about agents. For now, the companies’ revenue ambitions depend on a kind of access many users are not willing to grant. Big download numbers can start adoption, but they cannot substitute for confidence in what an agent will do with someone’s private life.
Daily AI news
Every day we pick what actually matters in AI and explain it plainly — no hype, no filler. Subscribe if you want to follow where the industry is going.
Only what matters — every day
Follow on X