Meta unveiled its entry into the AI assistant race earlier this month, amassing 5 million downloads in just 22 days and making it “the early frontrunner in consumer personal agents,” according to Forbes.
Meta’s site says that Muse “helps people stay on top of things, takes tasks and projects off their plate and turns long-term goals into action plans.”
Download the Straight Arrow app today to get the stories that matter free from manipulation, bias or agenda.™
Point phone camera here
But the rollout has seen some challenges — starting with music-fan outrage that the British band of the same name changed their long-held instagram and X handles with more than 5 million total followers to make way for the app.
On the product side, The Wall Street Journal called Muse “helpful and scary,” and emphasized its demand for wide user permissions to calendars, inboxes, and login credentials. ZDCNET described it as “the worst AI agent for privacy.” A hacker told ArsTechnica that a (now-patched) vulnerability allowed him to “manipulate the agent and leverage its privileges to do whatever we want,” and documents seen by Reuters suggest that the beta launch of its integrated phone-calling product is actually staffed by humans.
What else have we learned during the first month of Muse?
AI with a baby face
Muse is embodied by an extra-adorable little guy called Jolly and can be accessed on a Tamagotchi-like keychain screen.
Independent tech journalist Becca Caddy has called this design choice an “AI Labubu,” and suggests the company is employing a technique called “Kindchenschema” — using baby features to gain trust. In other words: cute-washing.
That adorable factor serves a key purpose in helping dispel potential trust concerns, according to Caddy.
Customers might decide that the cartoon makes the product seem safer. “Look at this sweet little creature helping organize my life! She just wants access to my bank account to help me save for a holiday,” Caddy wrote. “And she’s a cat! And she’s waving!”
This is likely a growing industry strategy, judging by OpenAI’s Tuesday announcement of their own agent product, “Dots”. The various “dots” characters feature bright colors, cute eyes and a very non-threatening beret.
READ MORE: Meta suddenly retracts new ‘gross’ AI feature. Is this customer-responsive or damage control?
The app can read your iMessages
Jason Aten, a reporter at Inc Magazine, said that Muse read his private messages with his editor and started recommending articles he could write, despite him having refused access. The app claimed it was reading the text of his MacBook notifications, but, in a screenshot Aten posted on threads, it appears to have accessed 187,000 rows of iMessage data stored on the laptop.
Meta has refuted the reporter’s claim that this happened without consent, but acknowledged that Muse can access all of a user’s iMessages if granted “Full Disk Access.”
ArsTechnica described some of the access permissions required by the app as undermining security systems that “Apple has spent years developing.”
That might sound like a familiar concern. A recent ChatGPT-MacBook integration, announced in August, does the same thing. And that product announcement prompted a privacy expert to tell Straight Arrow he couldn’t “think of anything worse.”
READ MORE: ‘I can’t think of anything worse than this.’ Privacy at risk as ChatGPT now connects to iMessage
Spilling over into IRL
Muse is also having real-world impacts.
Meta seems to be leaning into the fact that one of Facebook’s major draws is now Marketplace, positioning Muse as an agent that helps with tasks like online shopping and messaging about purchases.
Instagram is currently serving an ad with Muse (cutely adorned with a pink scarf and sunglasses) offering to be a “style scout.” But it might be wise to wait, given reports that the app recently gave out a YouTuber’s home address to a Facebook Marketplace buyer.
“Then they showed up without it even telling me until late tonight that it messed up,” the YouTuber, Matt Robb, posted on Threads. The AI, speaking for Robb, wrongly informed the buyer that he was home while they were outside.
Meta has responded, offering to help work out what happened. Ultimately, it became clear that the permissions required for the app to operate autonomously (“allow always”) meant that it decided to share the address that it had learned from its previous interactions with Robb.
Amazon has also raised privacy concerns about the app. As GeekWire reported, “it appears to capture and store customer credentials, which the company says could create privacy and security risks.”
This follows earlier reports that Meta AI — a different product that relies on the same models — identified a woman’s location from an old photo she thought she had deleted.
Round out your reading
- He built his own Flock-style camera to track police. The police didn’t like it.
- Trump’s apparent joke about letting Iran bomb LA and San Diego catches bipartisan condemnation.
- Grandparents are fighting for access to their grandchildren. Here’s what the law says.
- When the ‘victim’ is made with AI, child pornography cases get complicated.
- Fibermaxxing is trending. Here’s what the science says.
