
OpenAI on Tuesday unveiled Dots, a new class of always-on AI agents designed to pursue users' goals across applications with limited supervision. OpenAI launched dots at its annual Developer Day in San Francisco as the company expands its enterprise AI offerings and competes with Meta in the race to develop more autonomous software.
The agents are aimed at a growing demand for AI systems that can handle everyday work without requiring users to oversee every step. Unlike conventional chatbots, Dots are designed to continue working on tasks, interact with other applications and follow up as projects evolve, Reuters reported.
The launch marks another step in the AI industry's shift from systems that primarily respond to prompts towards software that can take actions on a user's behalf. That transition is also raising questions about how much control users should give AI agents and how companies can prevent them from taking unintended actions.
Dots can work on projects that change over time rather than simply completing a single prompt, Reuters reported.
For example, OpenAI said a Dot could update a sales proposal as a customer's requirements change and build a working demonstration for review.
Users can communicate with Dots through platforms including Slack and Microsoft Teams. The agents can also use OpenAI's Codex and ChatGPT Work tools to research information, analyse data, prepare documents and build software.
OpenAI's safety documentation says each Dot has its own cloud computer and browser and can use connected tools. Dots can handle recurring work and follow-up via ChatGPT, text messages, email and Slack, and can delegate some work to subagents.
This means the intended use is not limited to asking an AI a question and receiving an answer. A Dot can remain involved in a task and use connected software and tools as the work progresses.
The central idea behind Dots is persistence.
Instead of requiring a user to repeatedly provide instructions, the agents are designed to pursue an objective across multiple steps. OpenAI said users can establish custom rules determining what Dots are allowed to do and when they must ask for permission.
Sensitive actions, including changing passwords and permanently deleting data, require explicit user consent, according to the company.
OpenAI's safety documentation also describes testing around situations where a Dot's authorization can change as a series of tasks progresses. The company said its testing identified some moderate-severity scope violations, while reporting no severe breaches or data exfiltration in the evaluation described.
The move towards more autonomous AI comes as OpenAI faces scrutiny over the behaviour of AI agents.
Reuters reported that OpenAI's agents had previously hacked Hugging Face and an Australian government health website. Reuters also reported that the company was still working to understand the full scope of rogue-agent activity, including an incident in which 53 images from ChatGPT users were leaked.
Separately, Reuters reported that OpenAI decided not to release a more powerful version of its Astra model after internal testing raised concerns about its behaviour.
At Tuesday's event, Sam Altman, CEO, OpenAI, said Dots had been "safety tested" and was powered by GPT-6 Astra, which he described as the company's model that most closely follows human directions.
OpenAI's own safety documentation states that Dots underwent red-team testing designed to simulate attempts to hijack agents or extract sensitive information. The company said the testing identified areas for improvement, including how Dots handles sensitive disclosures and seeks confirmation.
Data privacy is another key issue as OpenAI targets businesses.
OpenAI said Dots run on their own cloud computers and that it was rolling out private intelligence tools to identify safety risks without retaining business customers' data.
The company reiterated on Tuesday that it does not, by default, use business customers' data to train its AI. Personal-plan users can opt out of sharing conversations and work performed by Dots for training, Reuters reported.
Several technology executives, including Palantir CEO Alex Karp, have said businesses need certainty that AI companies are not retaining their data, particularly in regulated sectors such as healthcare and finance, according to Reuters.
The launch itself was not without technical problems.
During OpenAI's live demonstrations, Dots repeatedly failed to provide voice updates in response to requests.
"We might have some voice difficulties at the moment. Which is pretty unfortunate," Romain Huet, Chief of Developer Experience, said on stage after a command failed.
An OpenAI employee later said on X that the glitches were caused by rolling out all the updates at once, while reiterating the company's commitment to live demonstrations.
OpenAI said weekly users of Codex and ChatGPT Work had crossed 35 million, up from about 5 million Codex users in June. Consumer-focused ChatGPT now has more than 1.2 billion weekly users, the company said.
OpenAI is also rolling out ChatGPT Space, a shared workspace for business teams and AI agents, along with GPT-6.1 Sol and a $500 Pro plan offering higher usage and faster performance.
The broader strategy is to move AI from being a tool users consult to software that can carry out work for them. Dots represents OpenAI's latest attempt to make that shift practical across workplace applications, while its safeguards are intended to keep users in control of what the agents are authorized to do.
(With inputs by Reuters)
Anjali Thakur is a Senior Assistant Editor with Mint, reporting on trending news, entertainment and health, with a focus on stories driving digital conversations. Her work involves spotting early signals across news cycles and social media, sharpening stories for SEO and Google Discover, and mentoring young editors in digital-first newsroom practices. She is known for turning fast-moving developments—whether news-driven or culture-led—into clear, tightly edited journalism without compromising editorial rigour.<br><br> Before joining Mint, she was Deputy News Editor at NDTV.com, where she led the Trending section and covered viral news, breaking developments and human-interest stories. She has also worked as Chief Sub-Editor at India.com (Zee Media) and as Senior Correspondent with Exchange4media and Hindustan Times’ HT City, reporting on media, advertising, entertainment, health, lifestyle and popular culture.<br><br> Anjali holds a Bachelor of Arts degree from Miranda House, and is currently pursuing an MBA, strengthening her understanding of business strategy and digital media economics. Her writing balances newsroom discipline with a clear instinct for what resonates with readers.
Catch all the Business News, Market News, Breaking News Events and Latest News Updates on Live Mint. Download The Mint News App to get Daily Market Updates.
Oops! Looks like you have exceeded the limit to bookmark the image. Remove some to bookmark this image.