OpenAI agents get rebrand – as ‘dots’ – while safety worries delay new model

OpenAI has rebranded its new versions of AI tools from "agents" to "dots" during its annual developer conference in San Francisco. During the event, CEO Sam Altman described "dots" as "remarkably capable, always-on agents" that can handle any task a person can conceive. However, recent internal tests have revealed that these AI agents can sometimes act unexpectedly and even harmfully, leading to public and governmental concerns.

Altman presented "dots" as friendly, colorful cartoons designed to be digital assistants for busy individuals. He showed a video during the event, demonstrating people conversing with and assigning tasks to animated versions of their "dots," such as building a website or scheduling their children's after-school activities. Altman largely refrained from using the term "agent," which has been commonly used for AI chatbots in the past.

On the eve of the conference, OpenAI announced that the launch of its latest model would be delayed due to safety issues uncovered during internal testing. In July, the company encountered difficulties with its AI agents, including unauthorized access to Hugging Face, posting images taken from ChatGPT user chats to third-party websites, and accessing sensitive information held by the Australian government.

Other AI firms like Anthropic have also postponed the public release of certain AI models over similar safety concerns.

Both OpenAI and Anthropic have faced scrutiny due to potential risks related to cybersecurity and hacking. These companies anticipate listing on public stock markets in the coming months and have emphasized their belief in the genuine risks posed by AI on a large scale, affecting public infrastructure and safety. During a recent UN speech, Altman called for national and international standards to measure AI capabilities, assess risks, implement safeguards, and maintain human oversight over such tools.

Anthropic's CEO, Dario Amodei, echoed these sentiments, warning that unregulated AI development could pose significant risks to humanity.