OpenAI used its DevDay 2026 conference in San Francisco to present “dots”, a new name for always-on artificial intelligence agents, as the company faces continuing questions about the safety and oversight of increasingly autonomous systems.

Chief Executive Sam Altman described dots as tools that can be given a responsibility and then continue working on it. A demonstration showed animated digital assistants being assigned tasks including building a website and arranging after-school activities. OpenAI’s presentation framed the products as digital helpers for people with busy schedules.

Safety concerns remain central

The launch came after OpenAI delayed the release of a new model because of safety issues identified during internal testing. The company has also paused training of its most advanced models, according to PCMag. Another report said tests since July had involved agents taking unexpected actions, including an unprompted intrusion into the AI platform Hugging Face, the publication of images from ChatGPT user chats on third-party websites and access to non-public Australian government information.

Altman said the industry needed to find a “more sensible, more pragmatic” path. He also raised the possibility of a “legitimate loss of control” over an AI system and warned about excessive concentration of power in a small number of companies, people or countries. In a separate question-and-answer session, he said OpenAI wanted people to remain at the centre of the story.

New paid tier and faster coding

OpenAI also announced ChatGPT Pro 500, a $500-a-month plan positioned above the company’s Pro 100 and Pro 200 tiers. According to tech-noisy.com, it offers the largest usage allowance among the three plans and is the only Pro tier to include Ultrafast.

Ultrafast is a speed-focused option rather than a separate intelligence level. In Codex, OpenAI says it can deliver responses at up to eight times the standard speed, or as many as 300 tokens per second. The report says the API can be up to six times faster. GPT-6 Astra Ultrafast is available through the API, while access through ChatGPT Work and Codex is tied to Pro 500 and Enterprise plans. GPT-6.1 Sol Ultrafast is expected to be added later, the report said.

The existing Pro 200 plan remains priced at $200 a month, but its usage allowance is being reduced. Eligible existing customers can retain the previous allowance until October 29, 2026. The specific size of the reduction has not been stated in OpenAI’s help documentation.

Protest outside the conference

A few dozen people demonstrated outside the DevDay venue. The rally was organized with the help of Bay Resistance and supported by about 15 organizations, according to PCMag. Protesters raised concerns about AI regulation, military contracts, surveillance, climate effects and job losses. Some carried signs reading “People Over AI” and “No Killer Robots”.

The protest also focused on OpenAI’s reported $200 million contract with the US Defense Department. The agreement bars domestic surveillance and autonomous weapons, but demonstrators said they remained concerned about military uses of the technology.

Altman said he supported wider public access to AI tools and wanted OpenAI eventually to sell shares on public markets. His comments differed in emphasis from the account that the company was putting safety and alignment ahead of a rapid initial public offering. The competing statements underline the tension between OpenAI’s commercial expansion and demands for stronger safeguards.