OpenAI Unveils "Dots" Agents and Paces AI Development for Safety Amid Math Controversy

OpenAI, the company behind ChatGPT, introduced its new autonomous agents called "Dots" during its annual DevDay 2026 conference in San Francisco. The company also used the occasion to reiterate its commitment to artificial intelligence safety, stating it would regulate the pace of its most advanced models when necessary. These announcements come weeks after a significant controversy: OpenAI's claim to have solved a century-old mathematical problem, Navier-Stokes, which generated both awe and skepticism within the scientific community.

New "Dots" Agents and the Pace of AI Development

The "Dots" are artificial intelligence agents designed, according to OpenAI, to be always-on personal assistants. Sarah Friar, OpenAI's Chief Financial Officer, told CNBC that these agents aim to help users be more productive and have a better day, acting as an assistant that can provide reminders or suggestions. These agents will be available across the company's paid tiers.

Beyond product announcements, safety was a central theme. Friar commented that OpenAI is being more vocal about its safety concerns because of what has been happening "out there in the world." In fact, the company has already held back its latest Astra model due to safety concerns. Friar stated that the company "will pace the frontier when necessary," indicating a willingness to slow down the development of advanced models to better address the technology's risks, a call the company had already endorsed earlier this month, as we reported in this note.

Regarding business, Friar highlighted an "incredible" third quarter for OpenAI, with 70% quarter-over-quarter growth and an enterprise business that doubled since July. She also mentioned that the company is "very well capitalized" after raising $122 billion in March, and that an Initial Public Offering (IPO) will happen "when the time is right for our business," without wanting to be distracted from its focus on safety and alignment. The CFO also defended the pricing structure of higher tiers, noting that customers are willing to pay more for the outcomes they achieve, according to CNBC.

The Navier-Stokes Problem Controversy

Weeks before DevDay, on September 9, OpenAI made a bold claim: it had solved the Navier-Stokes problem, a 90-year-old equation related to fluid movement. According to the company, 10,000 "coordinating agents" powered by its internal AI model arrived at the solution in 88 hours, using tools such as access to a cached version of the internet and the ability to run code, as reported by CNBC.

The Navier-Stokes problem is one of the seven Millennium Prize Problems, considered among the most difficult mathematical problems at the start of this century by the Clay Mathematics Institute of Cambridge, Massachusetts. Solving any of these problems comes with a one-million-dollar prize. So far, the institute has not commented on OpenAI's proposed solution, according to its website.

However, OpenAI's announcement quickly raised questions. Tristan Buckmaster, a professor of mathematics at New York University, expressed his doubts. Buckmaster and his colleague Levent Alpöge, who works for Anthropic (an OpenAI competitor), had been collaborating on mathematical problems, including Navier-Stokes. According to Buckmaster, Alpöge received "tips" that information about their progress had been passed to OpenAI. The mathematician noted that OpenAI's route to the solution was similar to their own work, a direction that is "not the direction one arrives at in a few days by giving a model the problem statement," according to his statement on his website cited by CNBC.

Buckmaster also questioned whether OpenAI models had been trained on, or had access to, their sessions in OpenAI's Codex, as he and Alpöge had used several large language models as part of their research. OpenAI responded in its statement that its effort to solve the problem began on September 1, after "hearing a rumor" about progress on the puzzle, which they later found out related to Alpöge and Buckmaster. The company assured that neither its researchers nor its agents saw the mathematicians' work until it was publicly released, and that no specific user data was accessed. However, they could not rule out that "de-identified data derived from their usage of our products helped improve our models," as cited by CNBC.

Why it matters

OpenAI's recent announcements show a company in full expansion, with new AI products seeking to integrate technology more deeply into users' daily lives. At the same time, the company faces increasing pressure regarding safety and ethics in artificial intelligence development. The decision to pace its most powerful models reflects a more cautious stance, aiming to balance innovation with responsibility. The controversy over the Navier-Stokes resolution, meanwhile, underscores the challenges and complexities that arise when AI ventures into cutting-edge scientific problems, highlighting the need for transparency and verification in a rapidly advancing field.