OpenAI will soon let ChatGPT users shop directly in chat by integrating Shopify’s native checkout flow, as indicated by newly uncovered code strings. This move follows similar in‑chat commerce features from Microsoft and Perplexity, and promises a seamless conversational shopping experience.

Red Hat advocates responsible AI through compact SLMs, optimized by Neural Magic’s sparsity and vLLM runtimes, and aligned via the InstructLab community project. This open‑source, hybrid‑cloud approach controls costs, preserves data sovereignty, and supports domain‑specific customization.

OpenAI’s latest reasoning models, o3 and o4‑mini, hallucinate up to 48% on PersonQA, outpacing older versions. Enterprises must temper performance benefits with robust validation and model selection strategies.

Huawei’s CloudMatrix 384 Supernode outperforms Nvidia’s 180 PFLOPs NVL72 with 300 PFLOPs, higher memory bandwidth, and sanction‑proof design, fueling China’s AI infrastructure ambitions.

Attackers weaponized the Gamma AI presentation tool in a living‑off‑trusted‑sites campaign, using multi‑stage redirects and real‑time credential validation. Behavior‑based email security and multi‑factor authentication are key mitigations.

OpenAI has released o3 and o4‑mini, its most advanced reasoning models yet, featuring integrated web, code, and image tools. Available today to ChatGPT Pro and Plus users, they pave the way for GPT‑5 and next‑gen AI agents.

Google Cloud Next 2025 unveiled practical AI agent use cases built for enterprise-scale. Explore how CIOs and CTOs can drive value with AI-powered search, automation, and intelligent development tools.

SAP’s SALT dataset gives AI researchers access to real, anonymized ERP data—closing the gap between generative AI research and real-world enterprise systems. It’s a foundational step in developing table-native enterprise AI.

Apple’s new AI strategy leans on synthetic and anonymized data to enhance email summarization, Genmoji, and writing tools — all while keeping user content private. Available in beta with iOS 18.5, it’s a bold privacy-first approach to AI training.

MIT researchers improved the PAC Privacy framework, making it computationally efficient and effective at protecting sensitive data while maintaining AI model accuracy. This method offers a better privacy-accuracy tradeoff, ideal for high-stakes industries like healthcare and finance.