OpenAI's Dots leans on privacy promises as it takes on Meta's Muse

OpenAI used its DevDay event to make privacy the centerpiece of its agent strategy, unveiling Dots with a pledge from CEO Sam Altman that the company wants to, in his words, set a new standard for privacy in frontier AI. According to The Verge, the announcement doubled as a running critique of Meta's Muse, the rival agent that launched a couple of months earlier with privacy commitments of its own — commitments that have since been tested by security flaws and awkward data-handling incidents.

Dots arrives with control as the selling point

As The Verge reports, OpenAI introduced Dots in late September and spent much of DevDay on trust. Alexander Embiricos, the company's Codex product lead, said onstage that OpenAI is focused on having the most trustworthy, safe and secure assistant. Altman demonstrated ways users can constrain their agents, such as setting a rule that bars Dots from making purchases above a certain dollar amount.

Executives also drew an explicit contrast with Meta. Glen Coates, OpenAI's head of app platform, argued that Meta does not have an AI product with 1.2 billion users, and said that launching something that makes those kinds of mistakes is something OpenAI would try to take the care to avoid. For business customers, the company presented a framework offering stronger data controls and zero data retention options, meaning data would not be stored on OpenAI's servers.

The Verge adds a caveat: Dots has so far avoided privacy scandals, but it is only available on ChatGPT subscription tiers starting at $100, which likely means far fewer people are using it.

Muse's privacy pitch has not held up

Muse launched as what Meta framed as a safer alternative to its predecessor OpenClaw, with CEO Mark Zuckerberg promising it was built from the ground up for privacy and security. Nat Friedman, head of product at Meta Superintelligence Labs, wrote on X that the goal was to build something like OpenClaw that Meta could make safe, secure, easy to use and scalable to billions of people. User data sits on a secure virtual machine, which Zuckerberg described as an isolated Linux computer with its own browser, CPU, memory and storage.

The product succeeded commercially — The Verge reports it topped the App Store charts and, per Apptopia, reached 600,000 daily active users in the US within weeks. The privacy record is murkier. Data is isolated between users, but Meta itself can still access it; a mechanism to cryptographically and verifiably prevent Meta from reading data in the VM is only planned for later this year. A security researcher exposed a zero-day vulnerability that could have let an attacker take control of Muse, since patched. According to 404 Media, several serious security issues surfaced shortly before launch, one of which could have let users reach Meta's own internal databases.

Muse also collects liberally. It defaults to letting Meta train models on user input, with an opt-out available. An Inc. reporter said Muse uploaded and read his private messages without being asked, and a YouTuber said it offered his address to a stranger via Marketplace — in both cases, per The Verge, the agent was apparently functioning as intended. Wired reported that the platform builds detailed profiles of users' friends and family.

Trust is the bottleneck for agents

The competitive logic is simple: agents need deep access to personal data to be useful, so labs are competing over who can be trusted with it. The Verge's Allison Johnson described hesitating when a bot asked for her bank information during a task, something Muse instead handles through a Stripe integration. Not every company is even making the promise — The Verge notes that Instinct was publicly criticized for terms of service that granted it broad access to user data, though it appears to have made adjustments since.

As The Verge characterizes it, the industry's playbook for popularizing agents has three parts: make them useful, make them cute and disarming enough to offset the creepiness, and make promises about privacy — then hope they hold up.

Why it matters

Privacy has shifted from a compliance afterthought to the primary competitive axis for consumer AI agents, and that is a meaningful change for the industry. Agents can only deliver on their promise if users hand over banking details, message histories and browsing sessions, and the early evidence — Muse's zero-day, its near-miss with internal databases, its training-by-default setting — shows how easily privacy commitments slip in practice. If OpenAI's enterprise controls and Meta's planned cryptographic guarantees actually materialize, users gain verifiable protections rather than marketing claims. If they do not, the industry's fastest-growing product category will rest on familiar data hoarding with a friendlier face.