Recent AI media releases, with links to original sources.
Only media releases directly from companies themselves or government departments are referenced on these pages.
A Microsoft Discovery Blog post describes using Microsoft Discovery AI tools to learn and practice RISC-V chip design, drawing on the author's satellite and semiconductor experience. It explains how AI-assisted workflows can lower barriers for engineers entering semiconductor design.
Microsoft shares lessons from acting as Customer Zero in its own AI transformation, aiming to become a Frontier Firm that is human-led and AI-enabled. The company reports measurable gains such as higher deal close rates and shorter supply-chain cycles, and outlines how councils, growth mindset, and Copilot adoption are reshaping work.
OpenAI is publishing a framework for tracking, investigating, and disclosing model misalignment, with six initial reports of unexpected model behavior. The framework aims to speed public disclosure even when investigations are incomplete.
Scale AI argues enterprises need deliberate red teaming of AI systems, not just prompt-level tests. The post explains how models can pass isolated checks while surrounding systems still fail under adversarial pressure.
Mistral and Mozilla are powering Firefox Smart Window beta with Mistral models for open, private, multilingual AI browsing. The collaboration aims to keep assistant features useful while respecting user privacy.
Cohere and OpenText announced a partnership to help governments and regulated industries move agentic AI from pilot to production. The deal combines OpenText enterprise data and context with Cohere North platform and enterprise models.
Lambda reports MLPerf Inference v6.1 results featuring the first agentic datacenter workload and the first model over a trillion parameters in the suite. The company also cites higher throughput on its benchmarked systems.
Live Nation and Salesforce announced an expansion of Agentforce to make show day easier for fans. The AI agents are intended to help with common concert-day questions and support at scale.
Salesforce is expanding Missionforce with purpose-built AI capabilities and a new OpenAI partnership for government missions. The updates target faster citizen service, less manual work, and support for always-on operations.
Salesforce explains why it post-trained its own CRM reasoning model instead of relying on one general model for every Agentforce task. Specialized reasoning is intended to improve intent detection, safety checks, and evaluation quality.
Researchers at the University of Manchester are using NVIDIA Earth-2 to forecast air pollution across the UK. The approach aims to deliver finer, more frequent air-quality insights than traditional chemistry models alone.
Emerald AI, Google, and NVIDIA launched an alliance to advance flexible, grid-aware AI data centers. The effort focuses on scaling AI factories while coordinating energy use with the broader power grid.
NVIDIA says its Vera Rubin NVL72 system delivered leading results in the MLPerf Inference v6.1 debut. The company argues system performance and software optimization are key to AI inference economics.
Toronto-based smartARM is using AI hardware and software to build more intuitive bionic prosthetics. Meta highlights the startup as an example of AI making everyday life more accessible.
Microsoft introduced Microsoft 365 G7, bringing productivity, AI agents, security, identity, compliance, and governance together for Government Community Cloud environments. It is positioned for agencies modernizing services under tighter cybersecurity and mission demands.
Microsoft outlines its multi-decade education work and how AI tools are being applied for students, teachers, and institutions. The post covers commitments around access, safety, and learning outcomes.
Google is rolling out agentic commerce updates aimed at helping merchants capture holiday shopping demand. The changes expand AI-assisted shopping experiences across Google Shopping products.
Google shares findings from research with RXN on how teenagers view and use AI today. The post outlines five takeaways for families, educators, and product teams designing for teens.
New OpenAI Economic Research describes how workers use AI for tasks outside their formal job descriptions. The findings highlight shifting work patterns as people adopt AI tools across roles and industries.
OpenAI is partnering with OATS to bring free, hands-on ChatGPT workshops to 1,000 older adults. The program focuses on practical everyday uses of AI for communication, learning, and staying connected.
OpenAI outlines how ChatGPT Work and Codex analytics in the Admin Console help teams link AI usage and spend to business outcomes. The guide covers usage views, task insights, Codex outcomes metrics, and Admin plugin and API reporting.
OpenAI is testing Sponsored Agents in ChatGPT Ads and rolling out AI creative tools in Ads Manager, with new HubSpot and Shopify integrations. The updates were published on September 16, 2026.
Grok Build now keeps notes on conventions, decisions, and project facts across sessions. Capture runs in the background after each turn; /memory browses the notes and /dream organizes them, and the feature is available now for new sessions.
Google is updating Gemini Notebook with real-time voice conversations on mobile, an audio recorder for lectures, interactive learning overviews, new quiz formats, and short video overviews. Eligible U.S. college students can claim one year of Google AI Pro, with Google AI Plus offers in other markets.
Google introduced Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, its latest live dialogue models for voice agents and natural conversation. The models are rolling out via the Gemini API, Google AI Studio, Search Live, Gemini Live, and enterprise previews, with SynthID audio watermarking.
Google shares four ways to use Gemini for household tasks such as meal planning, appliance troubleshooting, and home organization. The post ties consumer use of Gemini to findings from Google's AI & Economy ATLAS on AI outside the workplace.
Google outlines how its AI work is aimed at health, disaster resilience, learning, and economic opportunity, alongside a 300-language milestone. The article ties recent science releases and AI & Economy ATLAS insights to real-world impact.
Google Research describes using AI with the FireSat satellite program to detect early-stage wildfires as small as a car from space. The long-term goal is imaging the Earth about every 20 minutes to give responders more time.
Google says its technologies now support everyday interactions in more than 300 languages spoken by about 7 billion people. The post describes work to go beyond text translation toward models that handle tone, dialect, and community language needs.
Google released Gemini 3.8 Live, 3.8 Live Extended Thinking, and Gemini 3.5 Transcribe for developers via the Gemini API and Google AI Studio. The models target real-time speech-to-speech and speech-to-text product experiences.
Koa is trained with 27 years of Salesforce CRM intelligence to enable agents to reason through the complex, multistep tasks required for enterprise work. The collaboration also brings NVIDIA open models and accelerated computing into Missionforce for government and regulated organizations.
Siemens’ Teamcenter will work with Agentforce to bring engineering-grade answers directly into sales, service, and customer workflows. Agentforce qualifies every inbound lead for Siemens’ 18,000 sellers, converting inbound interest into productive sales conversations and revenue growth.
Expanded strategic partnership enables cross-platform agent reasoning and action on a shared infrastructure and data foundation, brings Salesforce workloads to Google Cloud through Hyperforce, and accelerates AI adoption. Today at Dreamforce 2026, Salesforce and Google Cloud expanded their strategic partnership to eliminate friction from fragmented AI stacks.
New integrations bring Salesforce data, context, and actions into Amazon Quick, AWS agents into Slack, and Agent2Agent (A2A) support for real-time, bidirectional voice between Agentforce Voice and Amazon Connect Customer. Amazon Web Services (AWS) and Salesforce today announced an expanded partnership grounded in trusted CRM data.
AIforce unlocks tremendous value by enabling agents anywhere to reason and take action across all the data, workflows, and logic inside Salesforce. Instead of fixed UI, AIforce empowers anyone to build composable, intelligent, live interfaces, wherever work happens.
HP pre-loads the Perplexity Windows app onto Windows PCs for sourced answers, local work, and connected apps. HP helped make the personal computer central to how people work.
Children’s Hospital of Philadelphia is using open source AI tools to model children’s hearts in seconds — with the goal of enabling safer, more precise care for kids with congenital heart disease. About 1% of all live births involve a congenital heart defect — and no two are alike.
New data visualizations make ATLAS data easier to explore and use, while new research provides insights on how scientists are using AI. How is AI being used by people every day, including in offices, labs, and creative studios around the world?
OpenAI describes how Fyxer built an AI executive assistant on OpenAI models, using more than 500,000 hours of EA workflows and real user feedback to draft replies in each person’s voice. The story covers specialised model routing for email context and Fyxer’s goal of managing more of users’ communication and coordination workload.
Recent Australian research shows that consumers are using AI to find information and compare options. SEENMETRIC has outlined five practical website changes that can help machines retrieve, understand and assess business information more reliably.
The TSA deployed Ace, an AI agent built with Salesforce Public Sector Solutions, to answer routine airport-security traveler questions. Ace handles about 100,000 conversations per month and resolves 96% without human escalation as part of a Salesforce Government Cloud modernization.
OpenAI describes how Perplexity uses GPT‑6 Astra to write communications, change software and monitor production systems with less human check-in than earlier models. The piece links stronger coding ability to better search and summarisation inside Perplexity’s answer engine.
Cognition is using GPT‑6 Astra to improve Devin’s ability to test software and demonstrate that it works for engineer review. Astra is being applied across Cognition’s product lineup, including demos such as testing an iPhone game with a recording and report.
OpenAI describes how its Habitat storage platform was adapted in Python to handle ChatGPT’s growth past one billion users. The post covers how Habitat supports fast, reliable access to product data at that scale.
Lambda outlines OpenResearcher, an open pipeline for training deep research agents that plan, search and gather evidence across many tool calls. The post focuses on the data bottleneck of capturing full research trajectories rather than simple Q&A pairs.
Runway shares research toward real-time video generation that optimises time-to-first-frame and streaming as you prompt. The approach moves beyond waiting for a finished clip after a single prompt.
ElevenLabs and Universal Music Group announced a multi-year licensing and product collaboration. Fans will get an ElevenLabs platform to remix and reinterpret tracks from participating artists, alongside joint AI audio tools for creators.
OpenAI and the GSA announced a multi-year deal giving federal, state, local and tribal governments free ChatGPT licence access and 50% off usage, plus expanded cyber-defender support. The offer extends last year’s federal programme more widely across the public sector.
Anthropic’s Threat Intelligence team reports disrupting malicious uses of Claude over eight months across cyber, influence, fraud and related harms. The September 2026 report shares case studies and how safeguards and partner sharing improved after each incident.
Google’s Gemini app is now available for Windows with an Alt+Space shortcut for desktop AI help without leaving your workflow. Users can draft documents, manage tasks and create images from the installed app.
