Oracle sets out four OpenAI-based offerings across its applications, security and small-business lines
Oracle announced four OpenAI-based offerings at Oracle AI World in Las Vegas on 27 October. Oracle AI Workspace is designed to connect Fusion Applications and NetSuite so employees can run approved workflows through ChatGPT plugins. Oracle AI Security Mesh, a managed service using OpenAI cyber models, helps customers find and fix vulnerabilities in their applications and infrastructure across OCI, other clouds and on-premises environments, with customers defining the testing scope. Oracle App Composer is meant to help employees build applications with Codex without coding expertise, with Oracle managing deployment and operations on OCI. An SMB bundle packages ChatGPT Business, Workspace, Codex, App Composer, OCI and NetSuite QuickStart, free to start with $12,000 of cloud credits a year (Oracle).
Oracle co-chief executive Clay Magouyrk said "OpenAI is transforming how enterprises solve business problems, from chat, to coding, to analytics, to autonomous agents". Greg Brockman, OpenAI's president and co-founder, said the expanded partnership "helps bring that opportunity to more companies, putting frontier AI into everyday operations". The release gives no availability dates and no prices beyond the SMB credits, and says it is not a commitment to deliver any functionality. A European buyer weighing AI Workspace should ask Oracle where the inference runs and under whose contract. The release states no data-residency and no EU-region terms.
Liquid AI puts six open-weight 2B models on one backbone
Liquid AI published LFM3 on 27 October: six models in the 2B parameter range branching from one shared backbone, covering text, vision, audio, omni and two decision variants that return calibrated answers. The weights are on Hugging Face, and the pretrained base, LFM3-1.7B-Base, is released for post-training. Its own figures put LFM3-1.7B at 38.42 on WebArena-Verified against 3.02 for its predecessor LFM2.5-2.6B, and LFM3-VL-2B's average across 62 vision benchmarks at 71.4 against 57.6 for LFM2.5-VL-3B (Hugging Face). The 1.73B text model takes 256K tokens of input and output together (model card).
The launch post calls the models open-weight without naming a licence; the model card for LFM3-1.7B carries the LFM Open License v1.0, not Apache 2.0 or MIT. A 2B omni model designed to run on a phone or a laptop keeps prompts, images and audio off a hosted endpoint. Read the licence before building on it: section 5(b) says commercial use by a legal entity whose annual revenue exceeds the $10m threshold "is not licensed under this Agreement" (Hugging Face).
Quick Hits
- Gartner says 77% of board members expect AI to deliver productivity and cost savings, yet only 3% of organisations exceed their cost-savings goals through productivity. Released 26 October at its HR Symposium/Xpo in Orlando (Gartner, The Register).
- UK ministers met representatives of 137 businesses about datacentres and no environmental groups. Spotlight on Corruption's "Zoned Out", published 27 October, counts meetings from July 2024 to June 2026, and 38 appointments of former ministers and senior civil servants by AI companies to September 2026 (The Register).
- Elvis rightsholders sued Anthropic over 274 songs. The complaint, filed 23 October in the Northern District of California, says Claude produces Elvis lyrics when prompted, and seeks damages and a share of profits, or $150,000 per work (The Register).
- London robotics startup Machenta came out of stealth with a $40m seed round co-led by Creandum and Union Square Ventures. Its model trains in simulation and, the company says, learns a physical task from one demonstration (Tech.eu).