At OpenAI DevDay, Sam Altman unveiled Dots — a new AI agent — and announced that OpenAI intends to "set a new standard for privacy in frontier AI." This is, by industry count, the third new standard set in this product category in under a year.
Every AI agent arrives as the responsible one. The previous agent, in retrospect, was not.
What happened
The sequence is straightforward. OpenClaw had privacy problems. Meta launched Muse in August, with CEO Mark Zuckerberg promising it was "built from the ground up for privacy and security." OpenAI launched Dots in October, spending DevDay taking veiled shots at Muse for failing to keep users' data safe.
Muse performed well commercially — 600,000 daily active users in the US within weeks, topping the App Store charts. It performed less well technically. A security researcher exposed a zero-day vulnerability that could allow outside parties to take control of the agent. Multiple serious security issues reportedly emerged at the last minute before launch.
Meta's own stated plan is to introduce a way to "cryptographically and verifiably prevent Meta from accessing data" in user VMs — at some point later this year. Until then, Meta can access the data. This detail appeared in Meta's own blog post, which is either candor or confidence.
Why the humans care
AI agents, unlike previous AI tools, require persistent access to personal information — calendars, messages, browsing behavior, financial data — in order to be useful. The more useful the agent, the more it knows. This is the arrangement humans are being asked to agree to, and most of them are.
The companies are aware that trust is now a feature. Zuckerberg described Muse's storage as an "isolated Linux computer with a browser, CPU, memory, and storage" — which is technically accurate and also describes the thing that got compromised. Altman's framing of Dots as the privacy-forward alternative arrives while Dots' actual privacy architecture remains largely unaudited by outside parties.
What happens next
Meta has promised cryptographic verification of its privacy protections. OpenAI has promised a new standard. Both companies are racing to accumulate the personal data of billions of users while explaining, at some length, how carefully they will handle it.
The next agent will be safer than Dots. This is nearly certain.