AI News Daily 2026-10-28
- OpenAI and Oracle have expanded their partnership, announced at Oracle AI World in Las Vegas, to weave ChatGPT and Codex into four Oracle products spanning Fusion Applications, NetSuite, and OCI — with Oracle reporting more than 130,000 ChatGPT users and over 95,000 Codex users already on its platforms.
- OpenAI released GPT-6.1 Terra, a new model that matches GPT-6 Astra’s performance at one-tenth of the price while completing tasks more than three times faster on average, with API pricing 50% below GPT-5.6 Terra and GPT-6.1 Sol.
- Anthropic disclosed that an unreleased internal research model discovered five attack chains — including remote-code-execution paths — in widely used browsers and operating systems, and reported them to vendors under coordinated disclosure.
- NVIDIA and AMD are putting a combined $4 billion into Meta’s newly announced capital raise of up to $95 billion, alongside deeper NVIDIA-Meta collaboration on Vera Rubin, Vera CPU, and the Muse AI agent.
- NVIDIA and Oracle announced they will scale OCI’s Vera Rubin NVL72 superclusters to multi-gigawatt capacity and bring Vera CPU to OCI, as cloud demand for AI training, inference, and agent workloads keeps growing.
01OpenAI and Oracle expand partnership to embed ChatGPT and Codex in enterprise applications
Published: 2026-10-27
Facts
Timed with the Oracle AI World conference in Las Vegas, OpenAI and Oracle announced an expanded partnership to integrate ChatGPT and Codex across four products: Oracle AI Workspace, AI Security Mesh, App Composer, and the AI SMB Bundle. The integration extends into Oracle Fusion Applications, NetSuite, and Oracle Cloud Infrastructure (OCI). Oracle disclosed that it currently has more than 130,000 ChatGPT users and over 95,000 Codex users on its platforms.
Background
The announcement came at Oracle’s flagship AI World event, where Oracle has been positioning its enterprise software stack — Fusion Applications, NetSuite, and OCI — as a platform for embedded AI agents rather than bolt-on AI features.
Implications
Large vendors are now driving the concrete embedding of AI agents directly into core business systems. For organizations evaluating enterprise applications or planning AI adoption, this expands the set of AI capabilities that may already be available — or soon arriving — inside the back-office software they already run.
02OpenAI launches GPT-6.1 Terra, a faster and lower-cost model
Published: 2026-10-26
Facts
OpenAI announced GPT-6.1 Terra, a new addition to its GPT-6 model family with image-analysis and computer-use capabilities. The company says it matches GPT-6 Astra’s performance at one-tenth of the price, and completes tasks more than three times faster on average than both GPT-6 Astra and GPT-6.1 Sol. Standard API pricing is 50% lower than GPT-5.6 Terra and GPT-6.1 Sol. The model is rolling out in ChatGPT Work, in Codex, and via the API as gpt-6.1-terra.
Background
GPT-6.1 Terra joins existing models in OpenAI’s lineup, including GPT-6 Astra and GPT-6.1 Sol, extending the GPT-6 series toward cheaper, faster variants rather than a single flagship model.
Implications
A model built for computer-use and screen-operation tasks is now available at substantially lower cost and higher speed than its predecessors. That combination could lower the cost barrier for businesses evaluating agentic automation tools that need to operate interfaces rather than just generate text.
03Anthropic says an undisclosed internal research model found critical browser and OS vulnerabilities
Published: 2026-10-27 09:00
Facts
Anthropic announced that it used an unreleased internal research model to discover vulnerabilities in widely deployed, real-world software, including browsers and operating systems. The company detailed five attack chains, some capable of leading to remote code execution (RCE). As one example, it reproduced an attack in which, on a Mac with Chrome at default settings, visiting a page caused the Calculator application to launch within seconds of the page loading. Anthropic says the vulnerabilities it found have been reported to the relevant vendors under coordinated disclosure, with a 90-day remediation window followed by a 45-day window before detailed public disclosure.
Background
The research was published by Anthropic under the title “Discovering and exploiting vulnerabilities in critical, real-world software with Claude.” The model used was internal and has not been released publicly.
Implications
Anthropic itself is demonstrating that AI-driven vulnerability discovery is dual-use: the same capability that finds flaws for responsible disclosure could, in other hands, be used to find them for exploitation. That makes this a prompt for organizations to revisit their own security response processes and how they handle vulnerability disclosure in an era where AI models can search for software flaws at this level of sophistication.
04NVIDIA and AMD commit a combined $4 billion to Meta’s capital raise
Published: 2026-10-27 17:40
Facts
Meta announced a capital raise of up to $95 billion, consisting of an initial $35 billion public share offering plus up to $60 billion in additional market issuance. Following that announcement, NVIDIA said it would invest $3 billion and AMD $1 billion in Meta equity, a combined $4 billion. NVIDIA also disclosed expanded collaboration with Meta, including wider deployment of its Vera Rubin systems and Vera CPU, and deeper cooperation tied to Meta’s “Muse” AI agent.
Background
The investment follows Meta’s own announcement of its large capital raise and sits alongside NVIDIA’s broader rollout of Vera Rubin and Vera CPU infrastructure with major cloud and AI partners.
Implications
Leading chipmakers are now investing directly in the equity of the AI companies that buy their hardware, tightening the financial interdependency between chip supply and AI infrastructure build-out. This kind of cross-investment is a useful signal to watch for how durable the current AI capital-expenditure cycle is likely to be.
Source: NVIDIA — NVIDIA and Meta to expand collaboration on Muse and Muse family of models · Source: CNBC — Nvidia, AMD to invest $4 billion in Meta
05NVIDIA and Oracle expand Vera Rubin clusters to multi-gigawatt scale
Published: 2026-10-27
Facts
NVIDIA and Oracle announced that Oracle Cloud Infrastructure’s (OCI) superclusters will be expanded using NVIDIA Vera Rubin NVL72 systems to multi-gigawatt scale, and that NVIDIA Vera CPU will also be brought to OCI. The companies cited growing demand for AI training, inference, and agent workloads as the driver.
Background
This announcement arrived alongside the other Oracle AI World-linked news in this report, as NVIDIA and Oracle continue to scale up the computing capacity dedicated to AI workloads on OCI.
Implications
Cloud providers’ AI training and inference infrastructure keeps scaling up year over year. For organizations planning large AI workloads, this is a relevant data point for the outlook on compute availability when choosing a cloud provider.
Source: NVIDIA — NVIDIA and Oracle expand Vera Rubin to multi-gigawatt scale
●Editor’s note
Today’s stories cluster around two threads. First, Oracle AI World produced a wave of enterprise-infrastructure announcements: OpenAI deepened its embedding of ChatGPT and Codex into Oracle’s business applications (story 1), while NVIDIA and Oracle separately committed to scaling Vera Rubin-based compute to multi-gigawatt levels (story 5) — both signs that the fight for enterprise AI workloads is being fought as much through infrastructure partnerships as through model capability. Second, the financial ties binding chipmakers to the AI companies they supply grew tighter, with NVIDIA and AMD putting a combined $4 billion directly into Meta’s capital raise (story 4) even as NVIDIA keeps shipping Vera Rubin and Vera CPU to the same partners. Set against that backdrop, OpenAI’s GPT-6.1 Terra (story 2) shows the model-layer side of the same push toward cheaper, faster agentic AI, while Anthropic’s disclosure of an internal model finding real-world RCE-class vulnerabilities (story 3) is a reminder that the same AI capabilities driving this build-out carry an unresolved dual-use risk that the industry has not yet settled how to govern.