Salesforce and NVIDIA introduce Koa, Google adds Gemini voice agents, and more

Hi there,

Welcome to today's AI Brief. Amazon launches Alexa+ in India with Hindi and Hinglish support, while OpenAI weighs a funding round at a $1.2 trillion valuation. Google introduces Gemini 3.8 Live models for voice agents, and D-Robotics closes a $400 million Series C round.

Salesforce and NVIDIA introduce the Koa CRM reasoning model, while Siemens and Salesforce expand their AI partnership for industrial sales and service. OpenAI, Anthropic, and Google DeepMind also discuss AI safety, as Elon Musk calls for cross lab AI safety testing.

Amazon launches Alexa+ in India with Hindi and Hinglish

Amazon launches its generative AI assistant, Alexa+, in India with English, Hindi and Hinglish support. The service can retain conversational context, complete multi step tasks and connect with smart home devices and partners including Swiggy, District by Zomato, MakeMyTrip and EazyDiner. Read more

OpenAI weighs funding round at $1.2 trillion valuation

OpenAI is reportedly discussing an investor initiated funding round that could value the company at about $1.2 trillion. The talks remain early stage, while CEO Sam Altman says the company will not pursue an IPO in 2026. Read more

D-Robotics closes $400 million Series C round

D-Robotics closed a $400 million Series C funding round on September 17, 2026. The company plans to expand its Sunrise chip portfolio and develop software for AI robot development. Read more

Google introduces Gemini models for real time voice agents

Google introduces Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking for voice conversations, visual inputs, parallel reasoning and tool use. The models are becoming available through the Gemini API, Google AI Studio, enterprise previews and selected Google products. Read more

Delos Data raises over $100 million for AI infrastructure platform

Delos Data raises more than $100 million from investors including Matrix, Playground and Capricorn's Technology Impact Fund. The Palo Alto company plans to expand its engineering, product and sales teams while launching its Nonstop AI infrastructure approach. Read more

OpenAI, Anthropic and Google DeepMind discuss AI safety

OpenAI, Anthropic and Google DeepMind have held discussions on potential AI safety standards, independent model evaluations and safeguards for frontier systems. The talks also raise questions about US antitrust rules, while OpenAI backs independent reviews under the FRONTIER Act. Read more

Salesforce and NVIDIA introduce Koa CRM reasoning model

Salesforce and NVIDIA introduce Koa, a CRM reasoning model for Agentforce built on NVIDIA Nemotron open models. The model uses synthetic scenarios based on Salesforce workflows to handle multi step sales, marketing, and customer service tasks within Salesforce infrastructure without training on customer data. Read more

Factory raises $200 million at a $5 billion valuation

Factory, a San Francisco AI startup, raises $200 million at a $5 billion valuation, more than triple its April valuation. The funding will support its AI agent platform for enterprise software development, testing and maintenance. Read more

Novo Nordisk partners with Anthropic for medicine development

Novo Nordisk will test Anthropic's Claude models and Claude Science in research and development workflows for medicine discovery and development. The drugmaker will also use the models for software development, with data governance and human oversight. Read more

TotalEnergies and Mistral invest €100 million in reservoir AI

TotalEnergies and Mistral launch a three year program worth more than €100 million to develop AI models for oil and gas reservoir work. A joint laboratory will combine TotalEnergies subsurface data and geoscience expertise with Mistral AI technology. Read more

Anthropic signs lease for Queensland data centre

Anthropic signs an agreement to use capacity at Zerra DC's planned 2.16 gigawatt data centre near Dalby, Queensland. The facility is expected to begin operating in 2027 and will support Claude inference, subject to Australian foreign investment and local council approvals. Read more

Canada and Germany commit up to CAD $300 million to LawZero

Canada and Germany will provide up to CAD $300 million to LawZero for its Scientist AI research, international hiring, computing capacity and a new Berlin office. The company also plans sovereign computing infrastructure in Canada with Hypertec and 5C. Read more

Meta introduces Meta One subscriptions with expanded AI access

Meta launches its global Meta One subscription service, offering Core and Premium plans with increased access to AI image and video generation, Instagram Restyle, and voice effects. Plans start at $14.99 per month, while the core Meta AI service remains free. Read more

Siemens and Salesforce expand AI partnership for industrial sales and service

Siemens and Salesforce will connect Teamcenter with Agentforce, bringing engineering data into industrial sales, service and customer workflows. Siemens also uses Agentforce to qualify more than 2,500 leads each month for 18,000 sellers in 132 countries. Read more

Elon Musk calls for AI labs and Chinese firms to test models

SpaceX CEO Elon Musk proposes that major US and Chinese AI companies test each other's models for safety before public release. He says rival labs have not agreed to the proposal. Read more

Crusoe and Perplexity partner on AI training and inference

Crusoe and Perplexity enter a multi year agreement covering AI model training and production inference. Crusoe will also provide Perplexity Enterprise Pro and Perplexity Max to its 1,800 employees. Read more

Mistral and Mozilla partner on Firefox AI browsing

Mistral models will power Firefox Smart Window, Mozilla's AI browsing assistant. The beta launches in France and North America, with privacy settings designed to avoid retaining user conversations by default. Read more

Anthropic combines Claude Cowork and chat in one interface

Anthropic combines Claude chat and Cowork, allowing the assistant to route tasks within a single interface. Claude Docs, Claude Slides and Claude Design will also become available in conversations, starting with Pro and Max users. Read more

Google gives AI agents access to Google Home through MCP

Google is rolling out early access to a Model Context Protocol server for Google Home. Compatible AI agents can control connected devices, view camera summaries and access smart home event history for eligible US Google Home Premium Advanced subscribers. Read more

Al Gore says AI safety risks outweigh data center emissions

In an interview with TechCrunch, Al Gore says electricity use from AI data centers is smaller than other sources of demand. He calls for greater focus on fossil fuel infrastructure and AI safety risks, while noting AI could help reduce global emissions. Read more

Hang Ten Systems adds $53 million to seed round

Hang Ten Systems, an AI software startup founded by former Infosys CEO Vishal Sikka, raises an additional $53 million led by Xora. The funding brings its seed total to $85 million and will support hiring as it helps enterprises build and modernize software with AI. Read more

Anthropic details plan for embedded third party AI safety evaluators

Anthropic proposes giving independent AI safety evaluators ongoing access to frontier AI companies, similar to employees. Evaluators would review training, assess alignment, report incidents and publish findings, with independence dependent on access and publication rights. Read more

SK Hynix discusses US RAM chip production with Intel

SK Hynix is discussing options with Intel to manufacture RAM chips in the United States, including possible space at Intel's planned Ohio factory. No agreement has been finalized, while SK Hynix continues construction of its $3.8 billion AI chip packaging and research facility in Indiana. Read more

Snap launches Specs Intelligence for $2,200 smart glasses

Snap introduces Specs Intelligence, an AI system for its Specs smart glasses and connected devices. The company also adds support for HBO Max and Spotify, announces enterprise partnerships with Amazon, Salesforce and Nvidia, and plans to ship the glasses later this fall. Read more

Icelandic Treble raises $18 million for voice simulation platform

Treble, an Icelandic voice simulation and synthetic data company, raises $18 million in a Series A extension led by Paladin Capital Group. Its tools support voice AI and audio hardware development, with Amazon and Logitech among its customers. Read more

Categories

Companies

Resources

Salesforce and NVIDIA introduce Koa, Google adds Gemini voice agents, and more - Daily AI Brief #541

We hope you enjoyed this article

Daily AI Brief