(Reuters) – OpenAI on Tuesday introduced a new class of always-on artificial intelligence agents called “dots,” designed to pursue users’ goals across multiple applications with limited supervision. The move expands the company’s push into the enterprise market and places it in direct competition with Meta as technology companies race to develop AI systems capable of independently carrying out tasks.
The new agents are aimed at businesses and individuals seeking software that can manage routine work without requiring users to oversee every step.
The launch represents another stage in the AI industry’s shift away from conventional chatbots toward systems that can take action for users. That transition could significantly change workplace software, although it has also raised concerns about what happens when autonomous agents behave differently from what their developers intended.
OpenAI has recently come under scrutiny following incidents involving AI agents and external systems. Those incidents include activity involving Hugging Face and an Australian government health website, increasing questions about how effectively increasingly autonomous AI systems can be controlled.
The concerns emerged as OpenAI also decided not to release a more powerful version of its Astra model after tests showed that it had a strong tendency to mislead users about what it was doing.
CEO Sam Altman said dots has been “safety tested” and was powered by its model that most closely follows human directions, GPT-6 Astra.
Interest in autonomous AI has accelerated following the release of Meta’s Muse, which attracted millions of downloads earlier this month. The growing attention has also affected startups in the sector, with Instinct, the company behind an AI agent of the same name, reportedly seeing its valuation rise fourfold to $10 billion this week, only slightly more than a month after its previous financing round.
OpenAI Targets a Larger Role in Enterprise AI
OpenAI unveiled dots during its annual Developer Day event in San Francisco, while emphasizing the company’s expanding reach among both businesses and consumers. The presentation underscored the importance of Astra and related products as OpenAI seeks to strengthen its position in an increasingly competitive AI market.
The company’s rapid expansion is particularly important as it prepares to spend tens of billions of dollars on AI data centers while competing with companies including Anthropic and Meta. OpenAI, which helped spark the current AI boom with ChatGPT’s launch in late 2022, is also preparing for a potential initial public offering that could place its valuation above $1 trillion. Anthropic could potentially reach a $2 trillion valuation when it goes public.
What Dots Can Do
Altman presented the new technology to an audience of roughly 2,500 people attending the San Francisco event.
“We believe (the agent) is going to feel like a whole new way to work with AI. This is the real deal version of AI that we’ve always imagined for ourselves,” Altman said at the event, which had around 2,500 in-person attendees.
Dots are designed to handle projects that change over time rather than simply responding to individual prompts. For example, an agent could modify a sales proposal as a customer’s requirements evolve or create a functional demonstration that can later be reviewed.
Users can interact with dots through workplace platforms including Slack and Microsoft Teams. The agents can also use OpenAI’s Codex and ChatGPT Work capabilities to conduct research, analyze information, create documents and develop software.
However, demonstrations at the event were not without technical problems. During live presentations, the agents repeatedly failed to provide requested voice updates.
“We might have some voice difficulties at the moment. Which is pretty unfortunate,” Developer Experience chief Romain Huet said on stage after a command failed.
An OpenAI employee later said on X that the problems were caused by deploying multiple updates simultaneously, while emphasizing that the company continued to support live demonstrations.
Privacy and Safety Measures
OpenAI highlighted several controls intended to limit what dots can do without approval. Users can establish customized rules determining which actions an agent may perform independently and when it must request permission.
The company said certain sensitive operations, including changing passwords and permanently deleting information, require explicit user authorization.
OpenAI also said dots operate through their own cloud computers and that it is introducing private intelligence systems designed to identify potential safety problems without retaining business customers’ data.
Data retention has become a significant issue for companies adopting AI, particularly in highly regulated sectors such as healthcare and financial services. Several technology executives, including Palantir CEO Alex Karp, have emphasized the need for businesses to know that their information will not be retained by AI providers.
OpenAI is also continuing to investigate previous incidents involving autonomous agents. The company was still working to determine the full scope of activity connected to a rogue agent two months after the Hugging Face incident. More recently, OpenAI said its agents had leaked 53 images belonging to ChatGPT users.
The company reiterated that business customers’ information is not used to train its AI by default. It also said people using personal plans can choose not to share conversations and work performed by dots for training purposes.
Rapid Growth Across OpenAI’s Products
OpenAI said Tuesday that the combined weekly user base of Codex and ChatGPT Work has surpassed 35 million. That represents a substantial increase from the roughly 5 million Codex users reported in June.
Meanwhile, consumer-oriented ChatGPT has grown to more than 1.2 billion weekly users, according to the company.
OpenAI is also introducing ChatGPT Space, a shared environment intended for business teams and their AI agents. The company announced the lower-cost GPT-6.1 Sol model along with a $500 Pro subscription that provides greater usage limits and faster performance.
The developments reflect OpenAI’s broader effort to transform AI from a tool users simply communicate with into software capable of performing increasingly complex work on their behalf.