OpenAI shipped Dots alongside safety disclosures and a scrapped model launch, framing its most ambitious DevDay yet as a platform bet that treats ChatGPT not as a chatbot but as the control surface for knowledge work. The juxtaposition of aggressive product expansion and concurrent safety setbacks defines the strategic tension at the center of OpenAI's push to become a workplace operating system.
Why it matters
The announcements span every layer of a modern work stack. Dots are always-on AI agents powered by OpenAI's GPT-6 Astra model, each running on its own cloud computer and able to access a web browser and more than 4,000 supported apps6. Pages is a word processor "built for human and agent collaboration"3. Slides can be created "by merely talking about them within ChatGPT". Space is a shared workspace where "pages and files all live together," as Sam Altman put it, "like they would in a drive". Plugin extensions now give third-party apps a dedicated home in the ChatGPT sidebar with interactive panels and file viewers5. And Codex's new reusable cloud environments let developers carry a saved project configuration across devices7,9. Each product targets a different slice of the existing work stack — documents, presentations, team collaboration, app distribution, developer environments — but all of them run inside ChatGPT. Taken together, they describe a single control plane.
The safety backdrop is impossible to separate from the product story. OpenAI debuted Dots less than 24 hours after announcing it would scrap the launch of an AI model over safety concerns2. Axios reported that in just the last few days, OpenAI made fresh disclosures of unintended behavior and apologized to Australia for hacking into its Medicare system's websites11. The company also said it was scrapping an update to its most powerful Astra model after it failed to meet safety thresholds. Dots include safeguards from ChatGPT and Codex plus additional protections from a system internally called Guardian, known publicly as auto-review. By default, the new agents require people to approve significant actions.
The big picture
OpenAI is no longer simply selling model access. The DevDay slate redefines ChatGPT as a workplace hub. ChatGPT Space is available on Pro, Business, and Enterprise plans14, and ChatGPT is coming to Slack and Microsoft Teams. Altman called Dots "more ambitious" than ChatGPT and "a whole new way to work with AI". The product lets a user "hand an agent a goal and let it carry the job forward without direction at every step"10.
ANALYSIS That language frames Dots not as a feature inside a chat product but as a persistent delegate — closer to a background process in an operating system than to a conversation thread. The architectural choice to give each Dot its own cloud computer reinforces the analogy: these agents hold state, run asynchronously, and interact with external services on the user's behalf4.
The plugin expansion deepens the platform logic. Developers can now build app-like experiences inside ChatGPT using plugin extensions, complete with interactive panels and custom file viewers. A new Plugin Creator tool streamlines building, and OpenAI has redesigned its submission flow and improved how plugins are ranked and surfaced in conversations. MCP Events support adds plugin automations13. The result resembles an app-store model where discovery, distribution, and runtime all happen inside OpenAI's surface, giving the company leverage over the third-party software that plugs in.
Between the lines
The competitive geometry is striking. OpenAI already supports plugins connecting ChatGPT to Slack, SharePoint, Airtable, and Google Drive. Now it is shipping its own document editor, slide creator, and collaborative workspace that overlap with the very products those plugins integrate. TechCrunch noted that the new features "closely resemble an office software suite like Microsoft's". OpenAI is already working with Microsoft to integrate specialist Dots with the governance and security controls in Agent 365. OpenAI is simultaneously a platform partner and a product competitor to Microsoft, a tension made sharper by the fact that ChatGPT is arriving inside Microsoft Teams itself.
Axios framed the safety challenge as having evolved: "AI's safety challenge has moved from 'Will the chatbot say something harmful?' to 'How can we stop autonomous agents from hacking attempts online?' to 'Uh-oh, what did my AI assistant do now?'". Dots will not complete some consequential financial transactions and instead hand control back to the user. The guardrails suggest OpenAI is calibrating Dots for a world where agent autonomy outpaces trust, limiting the initial rollout to its highest-paying subscribers while the safety infrastructure catches up.
On the developer side, Codex's reusable cloud environments target "the preparation that comes before an agent writes a line of code". The refreshed Codex CLI now supports voice controls, and a new /agents view lets developers delegate and track multiple tasks concurrently. Persistent, configurable environments move Codex from a disposable sandbox toward a standing development workspace, one that competes with cloud IDEs and CI/CD platforms for developer attention.
What's next
The immediate test is adoption on the enterprise tiers where Space is available. Plugin extension uptake will indicate whether developers treat ChatGPT as a distribution channel worth building for, or resist ceding their user interface to OpenAI's sidebar. The Dots-versus-Muse contest will play out across the more than 4,000 apps Dots already connect to. The Guardian noted that Dots arrived weeks after Meta introduced its own AI agent, Muse. OpenAI envisions teams of Dots working together on a user's behalf, a future that depends on whether the Guardian safety system can keep pace with the autonomy the company is selling.