OpenAI is promoting Dots, its new always-on AI assistant, as a proactive alternative to tools that wait for user instructions. But the product faces two significant challenges: the cost of running autonomous agents and concerns about the privacy implications of giving them access to personal data.

OpenAI brought its Dev Day roadshow to London this week, seeking to persuade the city's developers to adopt Dots. The company first announced the product at its annual developer event in San Francisco a month ago, describing it as its flagship agentic offering.

Unlike Codex, OpenAI's coding assistant, which waits for a user to provide instructions, Dots are designed to act proactively. The assistant learns a user's goals, monitors their inbox and work tools, and begins tasks before the user asks. Dots are not yet available in Europe, although OpenAI said it is working with European regulators to change that. How that regulatory engagement progresses is a key near-term development to watch for Dots' availability in Europe.

The AI industry is betting that proactive personal agents represent the next major stage of development, potentially creating as much value for consumers as coding agents have created for enterprise users. The market is becoming more urgent as companies including Apple and Goldman Sachs move to cap some employees' AI spending, a trend that could weigh on revenue growth at leading AI labs.

OpenAI is entering a crowded market. Meta's Muse launched first and has become highly popular with consumers. Elon Musk's xAI offers Grok Bot, while Anthropic also has its Duet assistant on the market.

OpenAI is positioning Dots by initially targeting high-end users with substantial spending power.

“There are many other assistants out there, some of which have great UI interfaces, but what they can accomplish can be less advanced,” Romain Huet, OpenAI's head of developer experience, said. “We really wanted to nail the enterprise and pro use case first, because I think that's where people will see the most value.”

OpenAI is rolling out Dots to enterprises and to its most expensive Pro subscribers in the United States.

The strategy also reflects the high cost of operating these agents. In a Fortune report published this week, Elon Musk said Grok Bot's running costs were “unsustainably high relative to revenue.” Adam Fry, who leads ChatGPT's consumer business, separately told reporters that Dots are “on the compute-heavy side.” The report is available at https://fortune.com/2026/11/03/musk-says-grok-bots-costs-are-unsustainably-high-while-openai-admits-dots-are-compute-heavy-is-the-personal-ai-agent-push-dead-on-arrival/.

OpenAI is attempting to reduce those costs by routing some Dot tasks to Terra, a smaller GPT-6.1 model, rather than sending every task to the larger GPT-6 Astra model. Huet said the company is focused on token efficiency. He said Terra can outperform Astra in some areas, including computer use and writing, at “a tenth of the price.” Because smaller models draw less compute per task, that routing choice directly shapes the cost base of a service designed to act continuously rather than only when prompted.

The company wants “the most amount of value delivered per dollar invested in these Dots,” Huet said.

Cost is not the only challenge facing personal agents. Privacy concerns have also emerged because these systems require deep access to users' data to provide assistance and take autonomous action. Critics, including Signal president Meredith Whittaker, have warned that the trade-off may carry substantial security and data-privacy risks.

Always-on agents also raise questions about consent. Whittaker has argued that people who email or message a personal-assistant user have not necessarily agreed to let an AI read and ingest their communications.

“Muse is logging people's contact lists, the people they talk to,” Whittaker said at the Masters of Scale Summit in October. “If you install Muse, and I am in your contact list, Muse now has my information. They are logging when we talk to each other…That could be extraordinarily serious for somebody working on human rights in an authoritarian country who is trying to communicate data with somebody outside of that country.”

In response to those concerns, Huet recommended that users build trust with personal agents gradually. Users could begin by connecting a calendar, then add Slack and Gmail. Huet said he has established limits for his own Dot.

“I personally said to my dot: I would love for you to categorize my email all the time. I would like for you to prep, possibly, what my answer might be, but don't send any emails on my behalf ever, and don't send any Slack messages on my behalf ever,” he said. “I want to do that myself.”

That approach leaves much of the privacy decision to individual users. During a later Q&A, OpenAI leaders said users can already see what their Dot accesses. They also said the company is considering “an easy way to pick and choose” which connections Dots can use. Whether that option ships, and in what form, is a concrete privacy-related development to watch.

The broader AI industry is also contending with infrastructure and spending pressures. Fortune reported this week that Big Tech's AI data-center problem is growing after the midterm elections: https://fortune.com/2026/11/04/voters-pull-the-plug-on-big-techs-data-center-boom/.

Fortune's upcoming Brainstorm AI conference will take place in San Francisco on Dec. 7 and 8. Speakers are scheduled to include OpenAI's Greg Brockman, Microsoft AI CEO Mustafa Suleyman, Palo Alto Networks CEO Nikesh Arora, AI researchers Fei-Fei Li and Andrew Ng, and LinkedIn co-founder and Greylock partner Reid Hoffman. The event information is available at https://conferences.fortune.com/event/brainstorm-ai-2026/HOME.

Beatrice Nolan is the author of the Fortune report. She can be reached at beatrice.nolan@fortune.com and on X at https://x.com/beafreyanolan.

This story was originally featured on Fortune.com: https://fortune.com/2026/11/05/openais-dots-are-coming-up-against-the-price-of-privacy-and-the-price-of-compute/.