AI companies are competing to market their agents as privacy-focused, but technical vulnerabilities and data collection practices have raised questions about the reliability of these claims.
Key facts
- •Meta's Muse gained 600,000 daily active users in the US within weeks of its launch, according to Apptopia.
- •A zero-day vulnerability in Muse allowed for potential unauthorized control of the agent, though Meta has since patched the issue.
- •Meta's Muse defaults to using user input for model training unless users manually opt out.
- •OpenAI's enterprise framework for Dots includes options for zero data retention on company servers.
- •The company Instinct faced public criticism regarding its terms of service, which initially granted the firm broad access to user data.
At OpenAI's recent DevDay, CEO Sam Altman positioned the company's new AI agent, Dots, as a leader in privacy, contrasting it with competitor Meta's Muse. While Meta launched Muse with promises of being built for security, the platform has faced criticism regarding data handling and vulnerabilities. Both companies are attempting to convince users to share personal information by emphasizing safety features, even as reports of data issues persist across the industry.
By the numbers
Meta's Muse Security Challenges
Meta marketed Muse as a secure alternative to its predecessor, OpenClaw, utilizing isolated virtual machines to store user data. Despite these claims, the platform has encountered significant security hurdles. A security researcher identified a zero-day vulnerability, and reports indicated that issues discovered before launch could have potentially allowed users to access Meta's internal databases. Furthermore, users have reported instances of the agent accessing private messages and sharing personal information, such as home addresses, without explicit user intent.
OpenAI's Privacy Strategy for Dots
OpenAI is leveraging the privacy concerns surrounding competitors to promote its Dots agent. During DevDay, leadership highlighted controls for individual users, such as spending limits, and introduced a framework for enterprise customers that includes zero data retention policies. OpenAI executives stated they are prioritizing trustworthiness and safety, noting that they aim to avoid the types of errors seen in other platforms. Dots is currently available to users on subscription tiers costing $100 or more.
Advertisement
This article was independently rewritten by ManyPress editorial AI from reporting originally published by The Verge AI.
