OpenAI unveiled Dots, an AI agent designed with configurable privacy controls and a no-data-retention option for enterprise use.
Meta launched Muse earlier and promotes it as built from the ground up for privacy and security, including data isolation in a secure VM.
Despite promises, experts warn that even with safeguards, real risks from misconfigurations and vulnerabilities remain.
Readers should know what controls exist and how to exercise them when using AI agents at work or home.
Quick read · 2 min
OpenAI unveiled its AI agent Dots with privacy controls and a potential no-data-retention option for some uses. Meta counters with Muse, a privacy-focused agent that stores data in a secure VM. Both teams say safety features reduce risk, but experts warn that misconfigurations or new vulnerabilities can still expose data. For everyday users, the takeaway is to check what data is collected, how it’s used, and what controls you can enable. Look for updates on cryptographic protections and enterprise options as these products roll out further.
Next up, watch for more specifics on retention policies, opt-out options, and how quickly vendors patch discovered flaws. If you’re considering using AI agents at work, ask about data handling in your organization and whether no-data-retention modes are available.
What you can do now: review privacy settings in any AI agent you use.
Ask about data usage for training and whether you can opt out.
Monitor for updates about security patches and feature changes.
AI labs are racing to win trust with promises of privacy as they roll out autonomous agents. OpenAI introduced its latest agent, Dots, at DevDay, highlighting controls that let users dictate how data is handled and even offer a no-data-retention option for some setups. Meta, meanwhile, pushed Muse as a safer alternative that operates with data isolated in what the company calls a secure VM. The idea behind both approaches is simple: give people more control over what happens to their information when they’re interacting with an AI agent.
But promises aren’t proof. A security researcher highlighted a zero‑day vulnerability tied to Muse before launch, and Meta has since patched it. Reports also point to ongoing questions about how much data is truly private, and whether certain features still enable data to be used for model training or other purposes. OpenAI says Dots will include strong enterprise controls and, in some cases, no data stored on OpenAI servers. Meta says Muse can and will make mistakes, but safety systems should limit the damage.
Muse is built from the ground up for privacy and security. Your data and credentials live on the Muse Secure VM — an isolated linux computer with a browser, CPU, memory, and storage.
In plain terms, the big question is who gets to see what you type into an AI agent and how that data is used. Both companies are trying to persuade users that their data stays separated from others and that you can set limits on retention and training. The risk isn’t just big data grabs; it can also be gaps in how features are implemented or mistakes that expose internal systems if a vulnerability is found.
02
How these privacy promises differ
OpenAI frames Dots as the most trustworthy, safe, and secure assistant, emphasizing guardrails for purchases, spending caps, and other rules you can set to keep a bot from acting beyond your wishes. Meta emphasizes data isolation inside Muse and a future path toward cryptographic and verifiable protections that would curb Meta’s access to user data in the VM, at least in theory.
03
What it could mean for everyday users
For ordinary users, the headline question is simple: should you use an AI agent at home or work, and how much privacy are you getting? The short answer is to look for clear, configurable privacy settings, understand how data could be used to improve models, and know what happens if you opt out of data collection. If you’re handling sensitive information, you’ll want to know whether the service offers no-data-retention options and whether it supports end-to-end controls for your team or household.
04
What to watch next
Expect more details about data retention policies, opt-out mechanisms, and the specifics of how each platform handles training data. Both companies have said they will provide enterprise-grade controls, but the exact terms can vary by product tier and region. Look for updates on how cryptographic protections will work in practice and what happens if a vulnerability is discovered after launch.
Dots is OpenAI’s new AI agent that includes configurable privacy controls and a possible no-data-retention option for certain deployments.
What about Muse from Meta?
Muse is positioned as a privacy-first agent with data isolation in a secure VM and safety measures designed to reduce risk from misuse.
Can these guarantees really protect my data?
Promises exist, but real-world protection also depends on how you configure features, how data is stored, and whether any vulnerabilities are found and patched quickly.
Drones are more common and capable than many people realize. Here’s a plain-language guide to what they are, how they function, and why recent drone strikes and wildfire work with them are making headlines.
