Oracle and OpenAI showed tools that let AI carry out approved work inside Oracle's systems, while the money behind the partnership got no time on stage.
ChatGPT and Codex are heading into Oracle's applications, data and cloud under an expanded partnership Oracle announced in Las Vegas on October 27, 2026. It brings OpenAI's models, the ChatGPT chatbot and the Codex developer tool into Oracle's applications, data and cloud. Oracle describes it as a collaboration to "redefine how the world's businesses run."
The work arrives as four new offerings aimed at employees, IT teams, security staff and founders. The pitch is simple: move from AI that answers questions to AI that carries out work across a company's systems. Keep an eye on OpenAI. It turns up on both sides of this deal, as the supplier of these tools and as a customer tied to a large share of Oracle's backlog.
“I think the new tools are sensibly fenced, but the partnership is a loop that only reassures as long as OpenAI keeps paying for Oracle's capacity.”
From answering questions to doing the work
Oracle says businesses still struggle to get AI into the heart of their connected systems, and that it can be the single place for that, one that centralizes access to the apps, data, cloud and controls companies depend on.
Its example starts with a supplier delay that puts a customer order at risk. Michelle Meng, an Oracle product marketing director, wrote that fixing it could mean finding a substitute product, checking inventory, confirming transport and updating the order. Oracle AI Workspace is meant to keep that goal in view while it coordinates the apps, agents, data and tools. Oracle's second example is the month-end close, when the finance team checks and locks the month's books. Here Workspace links three Oracle systems, Fusion Financials, EPM (Oracle's planning and reporting software) and NetSuite, and coordinates the approved work across them.
For builders, Oracle says AI already helps employees write apps, but deploying and running them takes extra expertise. In its example, staff build and approve a launch calendar in Codex, then publish it through Oracle App Composer as one shared app. App Composer scans the code for vulnerabilities and deploys it on managed cloud infrastructure. Oracle runs deployment and operations, while IT controls access, approved data, policies and costs, as Oracle's Blair Bozada put it.
All of this is Oracle's own description. What will the enterprise pieces cost, and when can a company switch them on? The sources don't say. And as Oracle itself notes, every new connection that lets an agent do more also gives a weakness elsewhere a longer reach.
An AI that tries the locks before burglars do
Oracle's answer is Oracle AI Security Mesh. Its example is an ordering app with a missing access check, the test that keeps each signed-in user to their own data. A customer logs in to view an order and can also pull up someone else's. Mesh maps the systems a customer has approved, tests attack paths and checks whether one customer can reach another's order. An attack path is the chain of steps an attacker might follow. A separate agent repeats the test to confirm the problem is real.
