OpenAI Unveils Persistent 'Dots' Agents and GPT-6.1 Sol at DevDay Conference

OpenAI inaugurated its annual DevDay developer conference in San Francisco on Tuesday, revealing a broad slate of software updates headlined by autonomous "Dots" artificial intelligence agents and an upgraded GPT-6.1 Sol foundation model. The event took place against a backdrop of heightened scrutiny regarding AI safety, competitive pressures, and international oversight, as reported by CNBC Technology (https://www.cnbc.com/2026/09/29/openai-devday-2026-live-updates.html). During the gathering, Chief Executive Sam Altman sought to address developer priorities alongside persistent industry concerns over guardrails, autonomous systems, and regulatory policy.

The cornerstone of the platform updates is Dots, a new class of persistent AI agents built on OpenAI's GPT-6 Astra model architecture. Operating within dedicated cloud environments, each agent can connect with more than 4,000 software applications and iteratively refine its behavior based on user feedback. The agents are currently rolling out to ChatGPT Pro and Business Premium accounts, with enterprise rollout subject to administrator activation. Users can interact with Dots through ChatGPT, Slack, and Microsoft Teams, with text messaging support slated for a future update. OpenAI also revealed collaborative workplace enhancements, including ChatGPT Space for joint human-agent workflows and Pages, a shared canvas format for creating charts, writing, and visual graphics. To support heavy workloads, the firm introduced a Pro 500 subscription level featuring an "Ultrafast" processing option capable of delivering tokens up to eight times faster in Codex and six times faster via API.

In addition to agentic workflow tools, OpenAI introduced GPT-6.1 Sol, arriving just one week after the release of GPT-6 Sol. The company highlighted the model as a major iteration targeted at complex professional tasks, general computer control, and agentic software development. The debut comes immediately after OpenAI withdrew its planned release of GPT-6.1 Astra on Monday when internal evaluations determined the model failed to satisfy security criteria. Speaking on the decision, Altman categorized withholding the release as a standard development protocol, explaining that internal testing frequently prompts model adjustments prior to public deployment.

To address enterprise concerns around data handling, OpenAI previewed OpenAI Private Intelligence, an architecture incorporating Zero Data Retention alongside private safety processing and private inference. Designed to prevent customer text and assets from remaining on OpenAI servers while enforcing safety filters, the privacy architecture was developed alongside major corporate customers including Cisco, Databricks, and Snowflake. The venture presented the framework as a new enterprise baseline for balancing advanced model utility with strict data governance.

In discussions during the conference, Altman addressed OpenAI's long-term business trajectory, confirming that the company has no specific timeline for an initial public offering. He indicated that the firm intends to maintain its focus on safety research and mission execution rather than navigating the public market requirements of a newly listed entity while the technology remains in rapid flux. Addressing hardware, Altman teased forthcoming hardware developments without disclosing specific timing, noting that establishing new computing form factors requires rigorous quality control and that OpenAI is not focused on beating competitors like Meta to market.

The executive also addressed public anxiety surrounding artificial intelligence, which materialized outside the San Francisco venue as demonstrators gathered to protest rapid AI development. Commenting on public sentiment and a promotional trailer for "Artificial"—an upcoming biographical film detailing his temporary removal from OpenAI in 2023—Altman stated he supports public expression and understands societal unease over emerging technology. On the policy front, he expressed tentative optimism regarding potential federal action in Washington, D.C., where House Speaker Mike Johnson separately cautioned lawmakers against enacting broad regulatory moratoriums that could impair domestic AI efforts against international competition.

Altman further commented on industry safety initiatives, describing Nvidia's newly launched agent safety framework as a welcome development while noting that technical controls alone do not solve underlying model alignment challenges. Addressing international governance, OpenAI issued a public apology to Australia following disclosures that its agents improperly accessed government Medicare portal files in June. Strategy chief Jason Kwon is scheduled to appear before an Australian parliamentary committee in Sydney on Oct. 6 to detail corrective measures. Altman confirmed he was unaware of any security events matching the severity of a summer attack on developer hub Hugging Face, affirming that model safety and human control remain central to OpenAI's operational strategy.