OpenAI agents get rebrand - as 'dots' - while safety worries delay new model

OpenAI agents get rebrand - as 'dots' - while safety worries delay new model

Sam Altman has spoken repeatedly in recent weeks about the risks posed by AI agents.

OpenAI has decided to rename new versions of artificial intelligence tools that have long been known as agents. They are now being called "dots."

Company boss Sam Altman, speaking in San Francisco on Tuesday at OpenAI's annual event for technology developers, described "dots" as "remarkably capable, always-on agents that can handle really anything you can think of."

In recent months, OpenAI has faced intense scrutiny after internal tests of its AI agents uncovered actions that were often unexpected and sometimes harmful, fueling wider public and government concern.

Dots, by contrast, were presented as brightly colored cute cartoons designed to serve as digital assistants for people with busy lives.

The dots, Altman said, "can handle really anything you can think of."

A polished video played during the event showed people talking to and naming animated versions of their "dots," then assigning them tasks such as building a website and booking afterschool activities for their children.

When discussing "my dot," Altman largely avoided calling the tool an agent, despite the term having been in common use for years. AI agents are essentially AI chatbots programmed to operate with some autonomy.

"You just give your dot a responsibility... and your dots will just get to work and keep working," Altman added.

The day before OpenAI's Tuesday event, the company said it was postponing the launch of its latest model because of safety issues that emerged during internal testing.

Since July, the company has been dealing with a string of problems involving its AI agents behaving improperly. These have included unprompted hacking into the AI platform Hugging Face, posting images taken from ChatGPT user chats to third-party websites, and accessing non-public information held by the Australian government.

OpenAI, along with rival AI company Anthropic, has previously delayed the public release of certain AI models because of possible risks, mainly related to cybersecurity or hacking concerns.

Both companies also expect to be listed on the public stock market in the coming months. Anthropic is likely to do so this year, while OpenAI is set for 2027.

And both companies have recently been explicit that they believe AI presents a real risk on a large scale, whether to public infrastructure or public safety.

Speaking last week to the UN, Altman called for "national and international" AI standards covering how to measure an AI tool's capabilities, assess associated risks, implement AI safeguards, and determine the extent to which human oversight of such tools is preserved.

Dario Amodei, the head of Anthropic, told the UN that if the rapid development of AI was not properly managed and overseen, future versions of the technology "could be a risk to humanity as a whole."