{"id":8488,"date":"2026-10-10T13:48:13","date_gmt":"2026-10-10T13:48:13","guid":{"rendered":"https:\/\/silvybrand.com\/?p=8488"},"modified":"2026-10-10T13:48:13","modified_gmt":"2026-10-10T13:48:13","slug":"privacy-ai-agent-promises-openai-meta-muse-dots","status":"publish","type":"post","link":"https:\/\/silvybrand.com\/?p=8488","title":{"rendered":"AI agent makers are promising privacy \u2014 will they deliver?"},"content":{"rendered":"


\n<\/p>\n

\n
\n

At this year\u2019s OpenAI DevDay, CEO Sam Altman unveiled the company\u2019s new AI agent Dots \u2014 and told the crowd that the company wants to \u201cset a new standard for privacy in frontier AI.\u201d OpenAI would spend the day taking veiled shots at Meta\u2019s Muse, its primary competitor, for failing to keep users\u2019 data safe. Yet Muse itself, a couple of months earlier, had launched as a supposedly safer alternative to predecessor OpenClaw \u2014 with CEO Mark Zuckerberg promising<\/a> it was \u201cbuilt from the ground up for privacy and security.\u201d<\/p>\n<\/div>\n

\n

In an age when companies hoard customers\u2019 personal data and cyberattacks are a dime a dozen, AI labs are trying to convince users to share even more information with their agents. Their latest strategy for success is one-upping the competition by promising that unlike their rivals, they\u2019ll keep user data away from prying eyes. The question is whether companies can keep these promises.<\/p>\n<\/div>\n

\n

Nat Friedman, head of product at Meta Superintelligence Labs, wrote<\/a> on X that the company\u2019s \u201cgoal with muse was to build something like openclaw that we could make safe and secure and easy to use and scale to billions of people.\u201d User data is stored on a secure VM, what Zuckerberg described as an \u201cisolated linux computer with a browser, CPU, memory, and storage.\u201d The company said<\/a> that most of the effort in building Muse was related to \u201ccareful design and engineering to operate [it] more safely.\u201d Meta\u2019s blog post continued, \u201cMuse can and will still make mistakes, but we expect they\u2019ll be much less frequent and cause much less damage due to the safety systems we\u2019ve built in.\u201d<\/p>\n<\/div>\n

\n

But though Muse topped the App Store charts and, per Apptopia, gained 600,000 daily active users in the US within weeks, Meta\u2019s promises seem to have fallen short. Although data is isolated from other users, and the company plans<\/a> to introduce a way \u201cto cryptographically and verifiably prevent Meta from accessing data in your VM\u201d later this year, Meta itself can still access the data. A security researcher quickly exposed<\/a> a zero-day vulnerability that could allow someone to take control of Muse (which has since been patched), raising fears of outside attacks. Multiple serious security issues reportedly cropped up at the last minute before launch<\/a>, one of which could have allowed users to access Meta\u2019s own internal databases, per 404 Media<\/em>.<\/p>\n<\/div>\n

\n

Muse also appears to liberally collect (and in some cases, hand out) data. It defaults to allowing Meta to train models on what users put into it, though it\u2019s possible to opt out. An Inc.<\/em> reporter complained that Muse uploaded and read his private messages<\/a> without him asking it to, and a YouTuber said it had offered his address to a stranger<\/a> via Marketplace \u2014 in both cases, Muse was functioning apparently as intended, but the user didn\u2019t realize how far it would go. The platform creates<\/a> \u201cdetailed profiles of all your friends and family,\u201d as Wired<\/em> reported. None of this is necessarily unexpected from Meta, but it certainly doesn\u2019t back up the idea that Muse is uniquely privacy- or security-conscious.<\/p>\n<\/div>\n

\n

OpenAI seized on this fact when announcing Dots in late September. Alexander Embiricos, OpenAI\u2019s Codex product lead, said onstage<\/a> at DevDay that OpenAI is focused on having the \u201cmost trustworthy, safe, and secure assistant,\u201d and CEO Sam Altman demonstrated ways that people could exert control over their individual Dots, such as setting a rule that it should never make a purchase over a certain dollar amount. \u201cI think we\u2019re in a different position to Meta in that they don\u2019t have an AI product that has 1.2 billion users,\u201d said<\/a> Glen Coates, OpenAI\u2019s head of app platform, adding that \u201claunching something that makes those kinds of mistakes is something that we would try to take the care to avoid.\u201d<\/p>\n<\/div>\n

\n

Altman himself also pushed the privacy angle<\/a> at DevDay. To court business customers, executives presented a framework allowing enterprises to have \u201cstronger controls\u201d over their data and zero data retention policy options (meaning no data is stored on OpenAI servers). Indeed, so far, there haven\u2019t been many privacy scandals with Dots \u2014 although since it\u2019s only available on the $100-and-up ChatGPT subscription tiers, there are also likely fewer people using it.<\/p>\n<\/div>\n

\n

People may still not feel comfortable giving OpenAI their personal data, in part simply because AI agents require so much of it. For instance, The Verge<\/em>\u2019s Allison Johnson felt uncomfortable typing in her bank information<\/a> when the bot asked as part of a relevant task (Muse has a Stripe integration to take care of that).<\/p>\n<\/div>\n

\n

Not all companies are making privacy promises, of course. Instinct was publicly criticized<\/a> for reportedly overly-broad terms of service that gave it unfettered access to data; since then, the company seems to have<\/a> made adjustments. For now, AI labs seem to be counting on a three-part approach to popularizing AI agent use with the general public: Make them useful, make them cute and disarming to help offset the creepiness, and make promises about privacy \u2014 and hope they hold up.<\/p>\n<\/div>\n

Follow topics and authors<\/strong> from this story to see more like this in your personalized homepage feed and to receive email updates.<\/span><\/p>\n