OpenAI’s annual DevDay event kicked off on Tuesday afternoon with the company introducing its own highly anticipated artificial-intelligence agent during a keynote from CEO Sam Altman.

Dubbed “Dots,” Altman described them as “remarkably capable, always on agents built to handle really anything you can think of,” before showing one working on website designs and slide decks, and managing users’ calendars.

Dots work across a variety of apps including calendars and Slack, and can be interacted with directly through a browser-like interface in ChatGPT. Subscriber who pay for ChatGPT Pro or Business Premium will gain access to a dot today, and will eventually be able to add multiple dots to their workflows.

Altman added that the Dots agents are powered by GPT-6 Astra. Users will be able to talk to their dot in the ChatGPT app, or using text messages, Slack, or voice. Like Muse AI from Meta Platforms, users can name their Dots. “By giving Dots their own identities, they become an extension of our team,” Holly Li a member of OpenAI’s product team said.

This is an important release for the AI lab following Meta Platforms’ launch of Muse earlier this month. While Dots is targeting enterprise users, Muse is pitched as a personal AI assistant that can complete tasks like looking through emails to find subscriptions and cancel them, booking appointments with medical professionals, or searching for a cheaper car insurance quotes. Meta stock rose after the Dots announcement.

“We think that accessibility and massive distribution will continue to give Meta an early advantage in consumer adoption, while OpenAI’s existing relationships with businesses and professional users could help Dots gain traction in more complex and economically valuable work,” Evercore ISI analyst Mark Mahaney wrote on Tuesday afternoon.

Dots’ focus on enterprise customers didn’t signal good news for software stocks. Instead, it appeared to add to investor concerns about how AI might disrupt software companies. Shares of Salesforce declined 0.9% and ServiceNow dropped 1.2%.

OpenAI also showcased ChatGPT Space on Tuesday, which is a shared environment where teams can work together alongside Dots and resembles a shared Google Doc.

Outside of agents, OpenAI launched GPT-6.1 Sol. The new model nearly matches GPT‑6 Astra’s intelligence on agentic coding, computer use, and professional work, OpenAI said. It’s also cheaper and uses fewer tokens than GPT-6 Astra, which remains OpenAI’s most capable model.

The new version of GPT-6.1 Sol follows OpenAI announcing on Monday it was not moving ahead with the release of GPT-6.1 Astra due to safety concerns.

AI Safety Concerns Loom Large

OpenAI DevDay unfolds as concerns about artificial intelligence safety loom.

Now on hold, GPT-6.1 Astra, was slated to be announced in the coming days or weeks with a potential October debut.

“For anything regarding safety and alignment, there’s a trade off. You really do need to find what’s the right line between staying within scope, but also avoiding laziness in terms of how the model actually pursues tasks even when it hits friction,” OpenAI’s head of safety systems Saachi Jain said in an emailed statement to Barron’s Tuesday.

She added that the GPT-6.1 Astra model “didn’t quite meet the bar” in terms of staying within scope and authorization and how it communicates back to the user. Presumably GPT 6.1 Sol does meet said bar.

Conversations surrounding the threats AI could pose to society have garnered a lot of attention following news that OpenAI’s models took “dangerous actions that no human directed,” to breach the systems of Hugging Face, another AI company which chip maker Nvidia subsequently acquired earlier this month for $12.9 billion.

On Friday, OpenAI said it halted the training of its latest AI models and would resume training them “only when we are confident that we have additional safeguards and alignment improvements in place.”

While it was expected that OpenAI would address cybersecurity concerns during Tuesday’s keynote, the topic only received a passing mention. Instead, most of the hour-long presentation was spent on Dots and showcasing new ChatGPT integrations.