OpenAI Unveils Dots Agent, Stays Silent on AI Safety

OpenAI CEO Sam Altman introduced a "remarkably capable, always-on" AI agent called Dots at the company's DevDay conference on Tuesday, and made no mention on stage of the security concerns that led OpenAI to withhold a newer model just one day earlier, according to CTV News's report, carried by The Associated Press.

Dots is designed to complete ongoing tasks proactively on a user's behalf, positioning it as a direct competitor to Meta's Muse, which we covered when Muse overtook ChatGPT in App Store downloads. According to CBS News's reporting, Dots can schedule meetings, make travel arrangements, handle budgeting, debug computer systems, and write software code. The agents run on GPT-6 Astra, the model OpenAI released earlier this month. Altman described the pitch on stage: "It's like an AI helper that always has your back, inspired by the cool versions of what we all watched in movies growing up."

The silence is the story here as much as the product. On Monday, one day before DevDay, OpenAI said it was holding off on releasing a planned successor to Astra, GPT-6.1 Astra, after researchers voiced safety concerns, according to Yahoo Finance's report. One outlet reported the model "failed to meet strict safety standards regarding scope and authorization." Altman did not address the scrapped model in his keynote.

Axios senior AI reporter Madison Mills told CBS News on Monday that the number of security incidents Anthropic and OpenAI are now investigating "could now number in the tens of thousands, potentially even more." That figure, if accurate, is dramatically larger than the specific incident counts we have tracked individually throughout the year, including the original Hugging Face breach involving roughly 1,200 agents and OpenAI's later disclosure of six additional misalignment incidents. Mills' figure comes from a single reporter's sourcing rather than an official company disclosure, so it should be treated as a claim worth confirming, not a settled count.

The timing here is deliberate and notable. OpenAI's announcements came as several of Altman's industry peers were meeting with President Trump the same day, including Dario Amodei of Anthropic, Amazon's Jeff Bezos, Nvidia's Jensen Huang, Elon Musk, and Microsoft's Satya Nadella, alongside senior administration officials and House Speaker Mike Johnson, according to the San Francisco Chronicle's reporting. Altman skipped that gathering to run DevDay himself, sending OpenAI president Greg Brockman in his place.

Rather than discussing safety, Altman used part of his keynote to argue for a different metaphor for the current AI boom. He said the best version of AI "is not about making people cogs in a giant machine, ever whirring faster and faster," instead framing the moment as closer to a renaissance than an industrial revolution. That framing lands directly alongside our earlier coverage of Anthropic's own IPO filing warning of catastrophic or existential risk, a genuinely different tone from a direct competitor filing similar public disclosures the same week.

For businesses evaluating OpenAI as a vendor, the gap between what the company disclosed on Monday, safety concerns serious enough to shelve a flagship model, and what it chose not to discuss in its highest-visibility public event a day later is worth noting directly when assessing how transparently OpenAI communicates risk to its own customers and developers.

For businesses considering deploying "always-on" proactive agents like Dots or Muse, the reported scale of ongoing safety investigations, potentially tens of thousands of incidents, is a reminder that these systems are being scrutinized by their own makers at a pace and volume most enterprise buyers do not see reflected in product marketing.

What is OpenAI's new Dots AI agent?
Dots is an always-on AI agent that proactively completes tasks like scheduling meetings, booking travel, budgeting, debugging code, and writing software, running on OpenAI's GPT-6 Astra model, unveiled at DevDay on September 29, 2026.

Why did OpenAI withhold a new model the day before DevDay?
OpenAI said Monday it was holding off on releasing GPT-6.1 Astra, the planned successor to its current model, after researchers raised safety concerns, with one report saying it failed to meet strict standards around scope and authorization.

Did Sam Altman address AI safety concerns at DevDay?
No. Altman made no reference on stage to the security concerns that led to withholding the new model, instead focusing his keynote on new product announcements and framing the AI boom as a renaissance rather than an industrial revolution.

OpenAI CEO Sam Altman unveiled Dots, a new always-on AI agent that can schedule meetings, book travel, and write code, at the company's DevDay conference, one day after OpenAI withheld a planned model, GPT-6.1 Astra, over safety concerns raised by its own researchers. Altman did not address the withheld model on stage, instead framing the AI boom as a renaissance, while OpenAI president Greg Brockman represented the company at a separate White House meeting with other AI industry leaders and President Trump the same day. A reporter's sourcing suggests the number of security incidents Anthropic and OpenAI are now investigating could number in the tens of thousands, a figure that has not been officially confirmed by either company.