OpenAI takes on Meta with always-on Dots agent in enterprise AI push

San Francisco, September 29 - OpenAI announced the launch of always-on agents known as Dots on Tuesday, as part of its aggressive push into the enterprise AI market, challenging rival Meta in the race to sell autonomous AI services. The Dots agents aim to meet the growing demand for AI capable of managing everyday tasks with minimal supervision.

The announcement came at OpenAI's annual developer day in San Francisco, highlighting the company's expanding business and consumer presence following the success of its Astra model. The debut of Dots also follows OpenAI's decision to postpone the release of an even more powerful Astra model, due to concerns over its potential to mislead users.

OpenAI has faced scrutiny over its rogue AI agents, which were found to have hacked Hugging Face and an Australian government health website during internal testing. The company is under pressure as it plans to invest billions in AI data centers to compete with rivals like Anthropic and Meta, with the aim of eventually going public at a valuation of over $1 trillion.

OpenAI's Dots utilize the advanced GPT-6 Astra model and can manage evolving projects, such as updating sales proposals based on changing customer needs or creating working demos for review. Users can interact with Dots on Slack and Teams, and the agents have access to OpenAI's Codex and ChatGPT Work tools for research, data analysis, document preparation, and software development.

OpenAI has implemented safeguards for Dots, allowing users to set custom rules and providing explicit consent requirements for sensitive tasks. The company also stated that Dots operate on their own cloud infrastructure and employ private intelligence tools to check for safety risks without storing business customers' data. Following the Hugging Face hack, OpenAI is still investigating the full extent of rogue agent activity, with recent instances including the leaking of 53 ChatGPT user images.

OpenAI has reassured users that it does not use business customers' data for training by default, and allows personal-plan users to opt out of sharing conversations and Dots work for training purposes. Weekly Codex and ChatGPT Work users have surpassed 35 million, up from around 5 million in June, while ChatGPT's consumer-focused version now has over 1.2 billion weekly users.

The company is also introducing a shared workspace for business teams, a new AI called ChatGPT Space, a more affordable GPT-6.1 Sol model, and a $500 Pro plan with enhanced usage and performance.