CourionAI Newsletter
← All news
openai 2 min read

OpenAI Launches ChatGPT Work as GPT-5.6 Finally Goes Public

GPT-5.6 is now available to everyone, and OpenAI is pairing the rollout with ChatGPT Work — an agent that plugs into your apps and works on projects for hours.

Risograph conveyor lines carrying paper sheets between floating window frames toward one finished document on a podium

Two weeks after GPT-5.6 launched under government-restricted access, OpenAI has gotten the green light for a full public rollout. And the company paired the moment with a new product: ChatGPT Work, an agent inside ChatGPT that connects to your everyday tools and works on complex projects for hours at a stretch, delivering finished results like a Word document or a spreadsheet.

The centerpiece is a new “Unified Plugins Directory” that bundles third-party integrations in one place. At launch it includes Google Drive, SharePoint, Slack, Microsoft Teams, Gmail, Outlook, Salesforce, Adobe, Zoom, LinkedIn, GitHub, Canva, and Dropbox. You can call a specific integration with an ”@” mention, or let ChatGPT figure out on its own which data source it needs. OpenAI’s example workflow: turn customer research into a campaign brief, generate the marketing assets, and adapt them for different markets — all from one prompt, with context carried across every step.

Rollout is staged. On web and mobile, Pro, Enterprise, and Edu accounts go first, with Plus and Business following in the coming days. Through the ChatGPT desktop app for Mac and Windows, though, it’s available right away on all plans, including free. Billing is usage-based: Work tasks draw from the same consumption pool as Codex, and how much a task costs depends on its size, complexity, and the model you pick.

What’s behind this? Under the hood, ChatGPT Work is built on Codex — the agent technology OpenAI has so far aimed mostly at programmers. In plain terms: the machinery that lets an AI work through a coding project step by step is being repackaged for office work. The Decoder notes that many of the pieces, like scheduled tasks and computer use, already existed in ChatGPT or Codex, so this is partly a rebrand — much like Anthropic turning Claude Code into Cowork for non-coders. It’s also a second run at plugins: OpenAI tried them in 2023 and co-founder Greg Brockman later admitted they “didn’t work at all because the models weren’t ready.” The bet is that GPT-5.6 finally is. A fair caveat: agents that act across your email and files raise real security questions. OpenAI points to an “Auto-Review” feature where stronger models check important actions first, and says it blocked 100 percent of data-extraction attempts in red-team testing — the company’s own numbers, worth remembering.

What this means for you: If you’re curious, the desktop app is the surprising shortcut — ChatGPT Work is live there for free accounts, so you can watch an agent draft a document from your own files without paying. If you already use AI at work, the interesting shift is usage-based billing: complex tasks cost more, so it pays to think about which jobs actually need an agent and which just need a chat.

Sources

Source: https://openai.com/index/chatgpt-for-your-most-ambitious-work/

Next story

OpenAI Says a Third of a Popular AI Coding Test Is Broken

OpenAI reviewed SWE-Bench Pro, one of the most-cited tests for AI coding skills, and found roughly 30 percent of its tasks flawed. It's pulling its endorsement.

Risograph wooden ruler cracked in the middle over a pile of exam sheets, with a magnifying glass hovering above the crack