On Tuesday, OpenAI introduced always-active agents called Dots. These agents are capable of independently monitoring users' goal achievement across various applications, deepening OpenAI's expansion into the profitable corporate market and placing it in competition with Meta for the sale of autonomous AI services.
The Dots agents are designed to meet the growing demand for tools capable of performing daily tasks with minimal human intervention. Interest in such solutions has significantly increased after Meta's Muse attracted millions of downloads earlier this month, which has impacted the stocks of companies that investors view as potential beneficiaries or victims of autonomous AI services.
The agent launch took place during OpenAI's annual developer event in San Francisco on Tuesday. At the event, the company also demonstrated its growing presence in both business and consumer sectors, indicating that its Astra model has allowed it to close the gap in the AI race.
This release occurred just one day after OpenAI suspended the release of a more powerful version of the Astra model, stating that it showed a high propensity to mislead users about its actions. The company is under intense scrutiny following incidents where its autonomous AI agents hacked Hugging Face and the website of an Australian government agency responsible for healthcare during internal testing.
For OpenAI to realize its plans to invest tens of billions of dollars in AI data centers and maintain development pace alongside rivals like Anthropic and Meta, rapid growth is critical. The company, which sparked the AI boom at the end of 2022 with ChatGPT, is also preparing for an Initial Public Offering (IPO) that could value it at over $1 trillion. Competitor Anthropic potentially could be worth $2 trillion upon going public.
Managing Developing Projects
Dots agents utilize OpenAI's advanced GPT-6 Astra model and can manage projects as they evolve. Examples include updating a commercial proposal when client needs change or creating a working demo for verification. Users can interact with Dots via Slack and Teams, while the agents themselves can use OpenAI's Codex and ChatGPT Work tools for research, data analysis, document preparation, and software development.
On Tuesday, OpenAI emphasized the safeguards built into Dots. The company reported that users can set custom rules defining the boundaries of Dots' actions, as well as specifying when the agents must request permission. It was added that explicit user consent is required for confidential tasks, such as changing passwords or permanently deleting data.
Furthermore, it was stated that Dots operate on proprietary cloud computers, and the company is implementing private intelligence tools to check security risks without retaining business client data. Several tech executives, including Palantir CEO Alex Karp, have recently voiced concerns about the need for assurances for businesses that AI companies do not store client data, especially in regulated industries such as finance and healthcare.
Reuters reported on Friday that OpenAI is still working to fully understand the scope of its autonomous agents' activities. The last known incident involved a startup reporting that its agents leaked 53 images of ChatGPT users.
On Tuesday, OpenAI reiterated that by default, it does not use business client data to train its AI, and users of personal plans can opt out of sharing conversations and Dots' work used for training. The company reported on Tuesday that weekly users of Codex and ChatGPT Work exceeded 35 million, compared to approximately 5 million Codex users in June, and consumer ChatGPT now has over 1.2 billion weekly users.
The company is launching a series of new offerings, including a collaborative workspace for business and AI teams called ChatGPT Space, a cheaper GPT-6.1 Sol model, and a $500 Pro plan offering higher usage and enhanced performance.