OpenAI Dots vs Meta Muse: Pricing, Privacy and Permissions

OpenAI Dots vs Meta Muse: what happens after the chat?

You ask AI for help. It gives you an answer. Then you copy the answer into an email, update the spreadsheet and chase the next step yourself.

Personal AI agents are trying to take over that last stretch. Meta launched Muse on September 8; OpenAI introduced Dots on September 29. Both give an assistant a cloud computer so it can keep working after the conversation ends. OpenAI's announcement and Meta's announcement

That's interesting. “Always on,” though, isn't much of a reason to pay. How much work can you stop watching? And how much time will you spend cleaning up when something goes wrong?

Those are the questions behind this OpenAI Dots vs Meta Muse comparison. Deepapi hasn't tested the two products head to head; the analysis below draws on their published documentation.

Dots, Muse and Muse Spark are different things

OpenAI Dots runs on GPT-6 Astra. Meta Muse uses Muse Spark. Muse is the agent product; Muse Spark is the underlying model. Mixing them up turns model capabilities, product features and subscription costs into one confusing comparison. OpenAI, Meta

The model matters. So do the apps the agent can reach, the context it remembers and what it does when a task stalls. A benchmark leaderboard can't settle all of that.

Our broader guide to autonomous AI tools looks at the same buying problem across the agent market: what can you actually delegate, and under what conditions?

Different starting points, increasingly similar work

Dots starts closer to ChatGPT, Codex and team workflows. OpenAI describes an assistant that follows evolving project context and brings work back for review. Dots announcement

Muse initially arrived through a standalone app and WhatsApp, with its own cloud environment for handling everyday tasks. Muse announcement

The easy shorthand is “Dots for work, Muse for life.” It already needs an update.

On September 29, Meta announced Muse for Small Business, including connections to tools such as Shopify, QuickBooks, Canva and Slack. Its announcement says publishing, sending and spending require approval. Muse for Small Business

Both products are moving toward the person juggling several apps who doesn't have time to narrate every step.

Think of a small shop trying to connect sales trends, customer feedback and social posts. Writing a punchy caption is the easy part. Finding the right data, deciding which change deserves attention and getting a useful draft to the right person take more work.

That's where the comparison gets interesting. A long connector list tells you about coverage. It doesn't tell you whether each connector supports the action you need.

Proactive doesn't mean permission to do everything

Dots draws a distinction that's easy to miss: proactive background research and user-assigned tasks have different permissions.

OpenAI says proactive research uses read-only tools. It can gather information from authorized sources, but can't message other people, change connected app content or control a browser. Follow-up actions still go through the usual rules and checks. Dots safety and privacy

Spotting something worth doing isn't the same as having permission to do it.

Muse's design notes emphasize activity records, editable memory and approval cards with explicit accept-or-decline controls. They also treat the question of whether an update deserves to interrupt you as a design problem. How Meta designed Muse

These sound like interface details. They determine how much supervision you're signing up for.

If an agent asks about every step, you're still directing the job remotely. If it quietly crosses a permission boundary, a few saved minutes may disappear into a much longer repair job.

Useful delegation means preparing the work and bringing back the decisions that need your judgment. Our Claude Gmail and Drive permissions guide examines a similar tradeoff once an assistant gets write access.

Read Meta Muse's privacy promises separately

Meta gives Muse an isolated virtual-machine environment and additional safety checks. The current Secure VM is different from the Confidential VM it plans to introduce later.

Meta's technical explanation says the current architecture restricts staff access through policy, but doesn't technically prevent Meta from accessing data when necessary to support, protect or operate the service. The planned Confidential VM would use cryptographic protections to further limit provider access. Meta's security architecture

That distinction belongs next to the privacy claim.

The same explanation separates advertising from training. Muse doesn't share conversations or VM data with Meta's advertising systems. Users can opt out of model training. Browsing external websites through the agent may still affect advertising indirectly. Meta's security explanation

“Not shared with ad systems,” “not used for training” and “the provider can't read it” answer different questions. Calling all three “fully private” would lose the conditions that matter.

Dots also has different training defaults for personal and enterprise accounts. Check the terms for the account you plan to connect. OpenAI's privacy explanation

For company data, our AI data-retention guide explains why retention and access policies belong in the buying decision.

Dots and Muse pricing: included isn't unlimited

Dots initially opens to eligible Pro and Business Premium users, with Enterprise access requiring administrator enablement. Availability varies by region and account rollout. The primary dot is included in eligible plans; that doesn't make it available to every ChatGPT subscriber. Getting started with your dot

OpenAI says chatting with a dot doesn't count against ChatGPT conversation limits. Codex and ChatGPT Work tasks it starts or manages still use their normal allowances. The extended deep-work allowance for the first month shouldn't be read as permanent unlimited usage. Dots launch details

Meta lists a usage-limited free version of Muse, Power at $20 a month and Maximum at $100 a month. The paid tiers list 500 million and 3 billion Muse tokens per week respectively. Benefits and availability can vary by region and account. Muse subscription details

Don't turn those Muse token allowances into an API price comparison without verifying what they measure.

Keep a more useful tally: allowance consumed per completed job, attempts required and time you spent intervening. A cheap subscription can still leave you doing most of the work.

Which one should you try?

If your work already revolves around ChatGPT, Codex and team tools, start by testing whether Dots reduces repeated explanations and task follow-up.

If you're starting with personal errands, or a small business using Meta business accounts and supported connectors, Muse may offer a more convenient entry point.

Those are judgments about product fit, not the results of a hands-on contest.

Pick a task you know how to check: summarize a week of customer feedback, prepare an event or turn business documents into an actionable checklist. Define what the agent may read, what it may change and which actions need your approval.

Then look at what comes back.

Did it finish? Did it miss a condition? Are you reviewing the result, or doing the job again?

That's where Dots and Muse will earn their place. The cute avatar can wait.

Editorial note: Facts checked October 3, 2026. This is a source-based comparison, not a hands-on review. Pricing, access and product capabilities may change.