Only 8% of US consumers would trust Meta with passwords, survey finds

The latest:

Meta’s new personal AI agent needs the kind of data access its own users are least willing to grant it. Muse, rolled out last week and described by Mark Zuckerberg as a personal AI agent built for everyone, can book doctor’s appointments, buy products, manage finances and send emails on a user’s behalf. Doing so requires entry to emails, calendars, shopping accounts and bank accounts, The Wall Street Journal reported.

Details:

• The survey: An Oppenheimer & Co. poll of US consumers found only 8% would trust Meta with their passwords, less than a third of the share willing to trust Google. Apple, ChatGPT, Claude and Grok also featured in the ranking, with a substantial share of respondents choosing none of the companies listed.
• The analyst view: Goldman Sachs analyst Eric Sheridan said in a note after the launch that consumer hesitancy around data privacy and security is likely the biggest potential headwind for broad-based adoption of consumer AI agents.
• Meta’s pitch: The company described Muse as safe, secure and private, saying it runs on a digitally walled-off virtual machine that no other agents can access. Zuckerberg said on X on Tuesday that release was delayed several months to focus on safety and security.
• The record: Meta paid a record $5 billion FTC fine over the Cambridge Analytica privacy breaches six years ago. Information on 533 million users leaked the following year, and transfers of European data to US servers drew a record $1.3 billion fine in 2023. User-safety litigation added to the damage, including a recent $18 billion settlement over harm to teenagers.
• The warning label: Meta concedes errors are coming. A notice shown at registration tells users the agent “may make mistakes or take unexpected actions, so review its work.” Because agents can buy, book and schedule autonomously, mistakes can cost users money directly.
• The advantages: Meta holds unparalleled social-media data, can plug agents into its Marketplace e-commerce platform, and is first mover in personal AI assistants, with no other major AI player yet offering a comparable product.
• The competition: Google is positioned to capitalize if the category takes off, pairing email, calendar and cloud-storage apps that could feed agents with a cutting-edge AI model.
• The economics: Light use is free; heavier users pay $20 or $100 a month. Oppenheimer estimated Meta would need 115 million subscribers at $20 a month to lift earnings per share roughly 20%, about double OpenAI’s paying ChatGPT base.
• The spending: Meta expects $137 billion of capital expenditures this year, funded by its fast-growing ad business. Subscriptions on Facebook, Instagram and WhatsApp, external access to its top models and a potential cloud push are other planned returns on that outlay.

Background:

Zuckerberg first outlined his vision for personal superintelligence last year and previewed Muse in an August essay, framing it as putting power in people’s hands to pursue their own aspirations. Meta counts 3.6 billion users.

Between the lines:

The Journal’s reading is that Meta’s first-mover lead cuts both ways: an agent empowered to spend money has a far lower error tolerance than a chatbot, and glitches during the rollout could push users away permanently. The subscriber math compounds the problem, since most consumers already paying for AI are customers of OpenAI and Google.

What’s next

Watch Muse sign-up and subscription disclosures against Oppenheimer’s 115 million threshold, whether Google launches a rival agent, and Meta’s next capital-expenditure guidance. Its stock is up about 10% since launch.