Evening Brief

Oracle and OpenAI bring ChatGPT and Codex to Oracle business software

Oracle says new tools put ChatGPT, Codex and OpenAI cyber models inside its apps and cloud, plus a bundle for small businesses.

  1. Oracle — Oracle reinvents enterprise computing for the AI era with OpenAI (Oct. 27, 2026)
  2. Microsoft AI — MAI Image 2.7 debut (Oct. 27, 2026)
  3. Google Cloud — Gemini Robotics ER 2 is now available on Google Cloud (Oct. 27, 2026)
Oracle’s Riverside campus in Austin, Texas, at night
Oracle’s campus in Austin, Texas. Photo: DronePhotographer / Wikimedia Commons (CC0)

Oracle says it is wiring OpenAI’s ChatGPT and Codex into its business software. Workers would ask ChatGPT for help, and plugins would reach into Oracle apps and data to get it done, the company says.

Oracle announced the deal Tuesday at its AI World conference in Las Vegas, in a release, and named four products.

What Oracle says it is building

Oracle AI Workspace uses plugins in ChatGPT to connect Oracle Fusion Applications, NetSuite, Oracle AI Database and Oracle Cloud Infrastructure, the release says. Employees can find information and complete approved workflows. Oracle says governance, approvals and controls stay in place.

Oracle AI Security Mesh is a managed service that uses OpenAI cyber models. Oracle says it helps customers find and fix flaws across its cloud, other clouds and on-premises systems. Customers set where testing is allowed. Oracle says it manages safeguards to cut the risk of outages.

Oracle App Composer lets employees build apps with Codex without coding expertise, the release says. Oracle handles deployment on its cloud. It also supports APEXlang, a new Oracle technology. Oracle says APEXlang avoids opaque code that is hard to maintain.

The Oracle AI SMB Bundle is free to start and comes with $12,000 a year in Oracle cloud credits, the release says. It packages ChatGPT Business, AI Workspace, Codex, App Composer and NetSuite QuickStart. Oracle says terms and restrictions apply.

“More businesses are choosing OpenAI as we make it easier to put AI to work through the tools and systems they already use,” OpenAI President Greg Brockman said in the release.

Oracle CEO Clay Magouyrk said the suite unites OpenAI’s AI “with Oracle applications, data, and cloud infrastructure.”

Our read: This puts ChatGPT in front of Oracle’s systems of record. That is where finance, payroll and supply data live. For buyers, the appeal is one chat window over many apps. The trade-off is depth of dependence. A company using AI Workspace, App Composer and the bundle ties its daily work to two vendors at once. Oracle says approvals and controls carry over, but the release does not say where prompts and data are processed or retained. The Security Mesh pitch is that customers need not build their own offensive testing tools. Customers would still be letting AI probe their own systems, with Oracle managing safeguards. The small-business credits lower the entry cost. They also start a firm on Oracle’s cloud from day one. Oracle’s own disclaimer says the plans are not a commitment to deliver and that timing and pricing may change. Everything here is a vendor claim until customers run it.

Microsoft’s new image editor

Microsoft released MAI Image 2.7, an image editing model, in preview, the company said in a post. Microsoft says it debuts at No. 3 on the Arena Edit leaderboard, which it says makes Microsoft AI the No. 2 lab there. Microsoft says it is No. 2 on Artificial Analysis’s image editing leaderboard.

Microsoft says the model improves on version 2.6 across all 10 use cases and seven editing actions that Artificial Analysis evaluates. Its biggest lead is in scene and style, where Microsoft says it scores 24 Elo ahead of the next-best model. The post does not name that model.

Google puts robot reasoning on its cloud

Google made Gemini Robotics ER 2 available in public preview on Google Cloud, according to a blog post by Carolina Parada, head of DeepMind Robotics, and Adrian Jones. Google calls the model a robot’s high-level brain. It plans tasks, checks progress on video and directs lower-level control models as tools, the post says.

Google says the model pinpoints the moment of a key event in video with 91.3% accuracy. It says the model reads more than 10 kinds of instruments, such as dials and thermometers. Developers can test behavior in simulators like MuJoCo-Warp and Nvidia Isaac Sim, the post says.

What’s next

Oracle’s release says the timing and pricing of the products may change. Microsoft’s model and Google’s robotics model are both in preview, the companies say.

Edition Oct. 27, 2026 (5pm) · Archive shelf