At OpenAI’s recent DevDay, CEO Sam Altman introduced Dots, the company’s new SI agent, and declared an intention to "set a new standard for privacy in frontier SI." The announcement positioned OpenAI in direct competition with Meta’s Muse, an earlier SI agent that had also been marketed on privacy grounds, raising questions about whether leading SI labs can deliver on their data safety promises.
What Happened
Meta’s Muse, launched a couple of months prior, was introduced by CEO Mark Zuckerberg as "built from the ground up for privacy and security." Nat Friedman, head of product at Meta Superintelligence Labs, stated the goal was to create a secure, scalable alternative to predecessor OpenClaw. The company described Muse as running on a secure virtual machine, characterized by Zuckerberg as an "isolated linux computer with a browser, CPU, memory, and storage." However, despite topping App Store charts and gaining 600,000 daily active users in the US within weeks, Muse faced significant security hurdles. A security researcher identified a zero-day vulnerability that could have allowed external control of the agent, which has since been patched. Reports from 404 Media indicated that last-minute security issues, including one potentially exposing Meta’s internal databases, emerged before launch. Additionally, users reported that Muse collected data aggressively, defaulting to allowing Meta to train models on user inputs and creating detailed profiles of friends and family.
OpenAI’s launch of Dots capitalized on these concerns. Alexander Embiricos, OpenAI’s Codex product lead, emphasized the focus on a "most trustworthy, safe, and secure assistant." Glen Coates, OpenAI’s head of app platform, noted that Meta lacks an SI product with 1.2 billion users, suggesting that mistakes in such a large-scale rollout are harder to avoid. At DevDay, Altman demonstrated user controls for Dots, such as setting purchase limits, and presented frameworks for enterprise clients offering "stronger controls" and zero data retention options. So far, Dots has avoided major privacy scandals, though it is currently limited to higher-tier ChatGPT subscriptions.
Why It Matters
The competition between OpenAI and Meta highlights a critical challenge for the SI industry: gaining public trust in autonomous agents that require extensive personal data to function. SI agents like Dots and Muse need access to emails, calendars, and financial information to perform tasks effectively, yet users remain hesitant to share this data. The incidents surrounding Muse, including a user’s private messages being read without explicit request and an address being shared with a stranger via Marketplace, illustrate the tension between utility and privacy. While Meta’s data isolation efforts are technically robust, the fact that Meta itself retains access to the data underscores the limitations of current privacy architectures in large-scale SI deployments. For developers and enterprises, the choice of SI agent now involves evaluating not just capability, but the verifiability of data handling and the robustness of security infrastructure.
The Bottom Line
SI labs are increasingly using privacy as a key differentiator for their agent products. While OpenAI’s Dots and Meta’s Muse both promise secure environments, the market is watching closely to see if these claims hold up under real-world usage and scrutiny. The industry’s success in popularizing SI agents may depend on balancing utility with transparent, verifiable data protection.