# OpenAI's Dots face the cost and privacy test

> OpenAI pitches its new proactive AI assistant Dots to high-spending users, but the technology faces mounting concerns about compute costs and data privacy.
- Title: OpenAI's Dots face the cost and privacy test · Fortune
- Summary: OpenAI pitches its new proactive AI assistant Dots to high-spending users, but the technology faces mounting concerns about compute costs and data privacy.
- Keywords: openai, dots, ai assistant, privacy, compute costs, anthropic, meta, xai, meredith whittaker, romain huet, grok bot, muse
- Source: Fortune — https://fortune.com/2026/11/05/openais-dots-are-coming-up-against-the-price-of-privacy-and-the-price-of-compute
- Author: Beatrice Nolan
- Read time: 3 min
- Topics: openai, dots, ai assistant, privacy, compute costs, anthropic, meta, xai, meredith whittaker, romain huet, grok bot, muse
## OpenAI brings Dots to London
OpenAI took its Dev Day roadshow to London this week to pitch developers on Dots, its new always-on AI assistant that learns user goals and starts tasks proactively.
## What Dots do differently
Unlike tools like Codex that wait for instructions, Dots watch a user's inbox and work tools, then start tasks before being asked. They are not yet available in Europe.
## The market for personal agents
The industry sees proactive personal agents as the next step for AI, potentially driving as much value on the consumer side as coding agents did on the enterprise side.
## Crowded competition
OpenAI enters a crowded market. Meta's Muse is already popular with consumers, Elon Musk's xAI has Grok Bot, and Anthropic has Duet on the market.
## OpenAI's strategy: high-end users first
OpenAI is rolling out Dots to enterprises and its most expensive Pro subscribers in the U.S., betting on high-spend users who will see the most value.

> We really wanted to nail the enterprise and pro use case first, because I think that's where people will see the most value.
## The compute cost problem
Dots are expensive to run. Elon Musk said Grok Bot's costs are unsustainably high relative to revenue, and Adam Fry said Dots are on the compute-heavy side.

> unsustainably high relative to revenue
## Routing tasks to cheaper models
OpenAI routes some Dot tasks to Terra, a smaller GPT-6.1 model, instead of sending them all to the larger GPT-6 Astra to manage costs.
## Token efficiency as cost control
Romain Huet, OpenAI's head of developer experience, said Terra can perform better than Astra in some areas at a tenth of the price, aiming for maximum value per dollar.

> a tenth of the price
## Privacy concerns emerge
Critics including Signal president Meredith Whittaker warn that personal agents need deep data access to be helpful, creating substantial security and privacy risks.
## The consent problem
Whittaker points out that people who email or message a Dot user never agreed to have an AI read and ingest what they send, raising consent issues.
## Real-world privacy risks
Whittaker warned at the Masters of Scale Summit in October that Muse logs contact lists and communication patterns, potentially endangering human rights workers in authoritarian countries.

> If you install Muse, and I am in your contact list, Muse now has my information.
## OpenAI's gradual trust approach
Romain Huet recommended users build trust with agents gradually, starting with a calendar, then adding Slack, then Gmail to limit data exposure.
## Huet's personal limits
Huet said he told his Dot to categorize emails and prep answers, but never to send emails or Slack messages on his behalf, keeping control himself.

> I want to do that myself.
## User controls still developing
OpenAI said users can already see what their Dot accesses and the company is thinking about an easy way to pick and choose which connections Dots can use.
## Privacy decisions left to users
OpenAI is leaving much of the privacy decision to individual users, offering gradual trust-building and connection controls but no mandatory safeguards.
## Key takeaway

OpenAI is rolling out Dots, a proactive AI assistant, to enterprises and Pro subscribers in the U.S., but high compute costs and privacy risks from deep data access pose significant challenges.