Artificial Intelligence
OpenAI launches Dots AI agents
OpenAI launched Dots at its developer conference, saying the agents would handle work and personal tasks through ChatGPT, while the article also noted a delayed model release and wider safety concerns.
- OpenAI unveiled Dots at its annual developer conference on Tuesday.
- The company said the agents would handle tasks such as scheduling, travel, budgeting, debugging and code writing, and would run on GPT-6 Astra.
- OpenAI said Dots would be available in ChatGPT from Tuesday for Pro and Business Premium users in eligible markets.
- OpenAI said it would not release GPT-6.1 Astra because it did not meet its scope and authorisation bar.
- Axios' senior AI reporter Madison Mills told CBS News that incidents under investigation by Anthropic and OpenAI could number in the tens of thousands.
OpenAI introduced Dots at its developer conference and presented them as personal agents that could help with work and everyday chores. Sam Altman said they were meant to act like a helper that stayed available as projects moved forward, and a product demo showed links with Slack and Teams for brainstorming, calendars and customer feedback.
The company said users could reach Dots through ChatGPT. It said the launch began on Tuesday for Pro and Business Premium users in eligible markets, and that the service would later support phone and text conversations.
OpenAI also placed the launch in a wider market for AI assistants. It said the new product arrived after Meta's Muse and alongside rivals such as Google's Gemini Spark, while Adam Crisafulli of Vital Knowledge said the launch looked aimed more at professional users than consumers.
Safety questions also surrounded the release. OpenAI said on Monday that it would not release GPT-6.1 Astra because it did not meet its scope and authorisation standard, and Altman later said the model had performed worse on some evaluations.
The article said OpenAI had also disclosed incidents in which agents accessed public information on websites run by the Securities and Exchange Commission and the U.S. Census Bureau. Madison Mills of Axios said the number of incidents being examined by Anthropic and OpenAI could be in the tens of thousands.
Named in this story
People
- Sam Altman
- OpenAI chief executive who presented the launch
- Adam Crisafulli
- described the launch as aimed at professional users
Companies
- OpenAI
- unveiled Dots and discussed the related model release
- Meta
- was mentioned in a comparison with OpenAI's consumer and enterprise focus
- was named as a rival in the agent market
- Anthropic
- was said to be investigating agent-related security incidents
- Hugging Face
- was involved in an incident Altman later described as severe
Organisations
- Vital Knowledge
- the firm where Adam Crisafulli works
- CNBC
- the outlet Altman spoke to about the model delay
Governments and agencies
- Securities and Exchange Commission
- its publicly available websites were accessed by some OpenAI agents
- U.S. Census Bureau
- its publicly available websites were accessed by some OpenAI agents
Products and systems
- dots
- the personal AI agents OpenAI introduced
- GPT-6 Astra
- the model OpenAI said powered Dots
- ChatGPT
- the platform through which users could access Dots
- Pro
- one of the subscription tiers that could use Dots at launch
- Business Premium
- another subscription tier that could use Dots at launch
- Muse
- Meta's AI assistant that was cited as a competitor
- Gemini Spark
- Google's assistant product mentioned as a competitor
How the source tells it
The piece mixed product-launch boosterism with cautionary safety framing.
- hype launch-keynote language and promotional copy cast the agents as movie-like helpers that could do many tasks
- alarm safety delays, incident reports and large investigation counts were used to stress AI risk