OpenAI's Dots and Meta's Muse duel over privacy promises

At OpenAI's annual DevDay conference, Sam Altman took the stage to introduce Dots, the company's new AI agent, and declared an ambitious goal: OpenAI intends to "set a new standard for privacy in frontier AI."

The remark, delivered in front of thousands of developers, framed privacy as a competitive battleground rather than a back-office concern. OpenAI spent the conference taking veiled shots at Meta's competing agent, Muse, accusing the rival of failing to keep user data safe.

The accusation landed awkwardly. Meta CEO Mark Zuckerberg had launched Muse only a couple of months earlier, marketing it as a safer alternative to its predecessor, OpenClaw. In a post on X, Zuckerberg told users Muse was "built from the ground up for privacy and security."

Two executives at the two most powerful AI labs in the West had each promised a privacy-first AI agent within a quarter of each other. The Verge's reporting captures the resulting collision.

What did OpenAI actually unveil at DevDay?

Dots is OpenAI's latest AI agent. Altman used the DevDay keynote to position Dots as the company's privacy flagship.

The phrase "frontier AI" is OpenAI's shorthand for the most capable models it ships, the systems the company considers capable enough to require the highest safety commitments. Extending that label to privacy was a deliberate move: Altman wanted Dots to inherit the trust status OpenAI has reserved for its largest models.

The exact technical mechanisms Dots uses to protect user data were not detailed in the opening keynote. OpenAI has not yet published a complete data handling policy for the agent.

Why is Meta in the crosshairs?

OpenAI's attacks on Meta were pointed but unnamed. OpenAI executives on stage accused the rival agent of unsafe practices without citing a specific incident or naming a specific vulnerability.

Meta had created that target by overpromising. Muse launched in mid-2025 as the successor to OpenClaw, an earlier Meta agent that had drawn user complaints about data handling. Zuckerberg's "built from the ground up" line was the public reset.

The post on X linked to the broader marketing campaign around Muse and committed Meta to a measurable standard. Each subsequent Muse release will be measured against that line.

What is the gap between marketing and mechanism?

The Verge's article highlights a recurring problem in the AI industry: privacy is now a marketing category, not a verified technical one. Companies hoard customer personal data as a default business practice. Cyberattacks hit major platforms regularly.

AI labs are trying to convince users, and increasingly regulators, that their products are an exception. The Verge's reporting questions whether the promises will translate into verifiable protections.

That question matters because AI agents are the first consumer software class to require deep, persistent access to personal data as a precondition for functioning. A calendar assistant that cannot read a calendar is useless.

How do the two launches compare?

The two launches share a structure. Both companies identified a privacy failure in their prior agent product. Both rebuilt the agent under a privacy-first banner. Both tied the launch to a CEO-level public commitment.

The differences are in scale and stage. OpenAI announced Dots at DevDay, the company's flagship developer conference, in front of an audience of professional builders who will integrate the agent into other products. Meta announced Muse through a social media post and a quieter blog entry, reaching a more general audience.

The DevDay venue gives OpenAI a structural advantage. Developers who build on Dots will encode its privacy promises into the apps they ship. Those apps will reach consumers who never heard Altman's keynote. The promises travel further when launched at a developer conference.

What should users and regulators watch?

Three signals will tell whether either company has earned the trust it claims:

• Published data retention, deletion, and access policies for the agent
• Independent third-party security audits of the agent's code and infrastructure
• Documented incident response procedures when a breach occurs

Neither OpenAI nor Meta has published a complete audit of Dots or Muse as of the latest reporting. The Verge's piece flags this absence as the central gap between promise and proof.

For regulators, the question is whether existing privacy frameworks cover agent-level access. Most current frameworks were written for static data collection, not for software that acts on a user's behalf.

What changes if the promises hold?

If OpenAI delivers on Altman's frontier-AI privacy standard, the competitive floor shifts. Every agent maker would have to match or exceed Dots's protections to remain credible. Meta, Anthropic, Google, and a long tail of startups would face pressure to publish comparable commitments.

If the promises do not hold, the damage concentrates on the loudest claimants. Trust in an AI product depends on the lab's reputation. A high-profile privacy failure at OpenAI or Meta would ripple through the entire agent category and likely draw regulatory response.

What does the timing tell us?

The two launches, separated by roughly two to three months, suggest the privacy-first agent is now a defined product category. Companies that ship an agent without a privacy commitment will be asked why.

That dynamic puts pressure on smaller competitors. Anthropic, Google, Mistral, and a long list of startups now have to either match the privacy posture set by OpenAI and Meta, or explain why their agents do not need the same protections.

Why this story matters

The Verge's piece is the latest sign that AI agent makers have decided to compete on trust, not just capability. Dots and Muse are both products that ask users to grant broad access to personal data. The companies now have to earn that access with more than a slogan.

OpenAI set its standard at DevDay. Meta set its standard on X. The two declarations will be measured against each other, and against the actual code both companies ship, for the rest of 2025 and into 2026.