DX Today AI Daily Brief: Saturday, September 12, 2026

0:00It's Saturday, September 12, 2026. You're listening to the DX Today AI Daily Brief. Today, Kohir and Nvidia both chase multi-billion dollar deals tied to the biggest names in AI, Salesforce pushes its agents further into the workplace. And a new Mexico court hands down a stock warning about chat GPT in the courtroom. Let's get into it. Canadian AI company Kohir is in advanced talks to raise as much as $3 billion in new funding, according to Bloomberg and the Globe and Mail. The round could value the company near $20 billion, roughly double what investors put on it earlier this year. Kohir has built its business selling large language models directly to banks, governments, and other enterprises that want to keep sensitive data in house. A pitch that has set it apart from consumer facing rivals. Terms are still being finalized and the amount could shift before any deal closes. If it goes through, the raise would rank among the largest private funding rounds in Canadian technology history.

1:03Speaking of big AI bets, Nvidia is in talks to investors much as $10 billion as an anchor backer in anthropics planned initial public offering, according to Reuters and Bloomberg. The chip maker's potential stake would come as anthropic, the maker of the Claude family of AI models, prepares for what could become one of the largest tech listings in years. Sources caution that terms remain fluid and nothing is finalized. An investment of this size would deepen Nvidia's ties to one of its biggest customers. Following a similar strategy, Nvidia has used with other leading AI labs. For anthropic, Nvidia's backing would send a strong signal to other prospective investors ahead of a listing that could value the company well into the hundreds of billions of dollars. Meanwhile, in Silicon Valley, Google has completed a talent and technology licensing arrangement

2:05worth roughly one and a half billion dollars with mechanize, a small but closely watched AI coding startup, according to Business Insider. The deal brings mechanizes researchers and its underlying agent technology into Google's fold. Though it stops short of an outright acquisition, mechanize had drawn attention in Silicon Valley for its ambitions to automate large swaths of software engineering work using autonomous coding agents for Google. The arrangement adds fresh talent to its push into agentic AI tools that can plan and execute multi-step coding tasks with minimal human oversight, an area where it faces intense competition from open AI, anthropic, and a wave of well-funded startups. Now, over to Washington. In Washington, Senate negotiators are weighing new legislation that would require AI companies to actively mitigate known major risks from their systems, according to Reuters. The proposal reportedly centers on a so-called duty of care standard, which could come with new disclosure requirements and give courts a role in reviewing

3:09where the companies met their obligations. The measure is still being negotiated and has not been finalized, with lawmakers from both parties working through disagreements over scope and enforcement. If it advances, it would mark one of the most direct federal efforts yet to hold AI developers accountable for catastrophic harms, an issue that has divided Congress for much of the year. From policy to robotics. Robotics startup Mecca AI is nearing a $500 million valuation in a funding round led by Sequoia Capital, according to TechCrunch. The company sits at the center of a broader rush among AI labs and robotics firms to secure high-quality training data for physical robots, a resource that remains far scarcer than the text and image data that trained today's leading language models. Investors are betting that whoever controls the best robot training data could have an edge as humanoid and industrial robots move from demonstrations toward real deployment.

4:10The round has not yet closed, and the final valuation could still shift before terms are complete. Halfway around the world. The United Arab Emirates is revising its plans for a massive AI data center campus, following recent Iranian attacks in the region, according to Reuters. Sources say the country may shift from a single concentrated campuses toward a more dispersed network of facilities, while also considering underground construction and stronger air defenses. The project's developer G42 set specifics remain under continuous review. The reported five gigawatt campus has been a centerpiece of the UAE's ambitions to become a global AI infrastructure hub, drawing partnerships with major US technology companies. The revised approach underscores how instability in the region is now shaping decisions about where and how to build the physical backbone of the AI boom. Back to deal making. Private equity firms, Francisco partners and KKR

5:10have agreed to buy minority stakes in team system, an Italian business software company that has leaned heavily into AI features for its enterprise customers, according to Reuters' exclusive. The deal could value team system at as much as 10 billion euros. The transaction would let existing owner Helmen and Friedman bring in new capital while retaining a significant position in the company, a structure increasingly common among private equity firms looking to hold on to their best performing assets longer. Team system has built AI tools into its accounting and business management software used by hundreds of thousands of small and mid-sized companies across Italy. Now a smaller story. Game creation platform Sunlab has integrated OpenAI's newest flagship model, GPT-6 Astra, into its Sunlab 2.0O platform,

6:11according to a company announcement. The update is designed to bring more advanced reasoning to the process of building playable games, letting users describe an idea in plain language and have the system handle more of the underlying design and coding work. Sunlab has positioned itself as a tool for people without traditional programming backgrounds who want to build browser-based games. And the company says the upgraded model improves the platform's ability to handle more complex game logic and interactions, speaking of fast-moving startups. AI startup discovery loop is seeking a new funding round at a valuation of around $50 billion, according to Business Insider, just weeks after the company emerged from stealth at a $10 billion valuation. The startup is linked to Jeff Dean, the longtime Google AI leader, and has drawn intense investor interest despite disclosing few details about its underlying technology. The rapid jump in valuation, if our version if it holds,

7:13would rank among the fastest markups for any AI startup this year. It also highlights how eager investors remain to bet on new ventures tied to proven AI research talent, even without a public product or a clear business model. From funding to the courtroom. New Mexico's highest court has sanctioned a defense attorney and fined him $5,000 after finding that a chat GPT-assisted legal brief in a murder appeal included fabricated witnesses and invented police testimony, according to Reuters. The court struck the flawed filings, barred the attorney from appearing before it while disciplinary proceedings continue, and referred the matter to the state's disciplinary board. The case is the latest in a growing string of court sanctions against lawyers who have relied on AI chatbots without verifying their output, a problem that judges across the country say persists, despite repeated warnings from the bench. Back to the enterprise. Salesforce has rolled out a new lineup of what it calls job-ready AI agents under its agent force

8:15platform, aimed at handling sales, service, and other business functions with less human setup than earlier versions required. The company also introduced a long, horizon agent runtime, designed to let an AI agent pursue a goal over days or weeks rather than completing a single task within one conversation. Salesforce is positioning the update as a direct answer to criticism that many enterprise AI agents still require heavy customization before they can handle real work, as it faces growing competition from Microsoft, Google, and a wave of startups. And finally to China. Chinese AI company Moonshot AI is targeting about $2 billion in annualized revenue by the end of the year, according to Bloomberg, as its Kimmy K-3 model gains traction with cost-conscious customers. The company has positioned Kimmy K-3 as a cheaper alternative to leading American models while still delivering competitive performance on many tasks, a strategy that has unsettled rivals

9:21in both China and the United States. Moonshot's projected revenue growth reflects intensifying price competition among Chinese AI labs, several of which have moved aggressively to undercut Western pricing on their models. Analysts say the coming months will show whether cheaper Chinese models can convert strong benchmark scores into lasting enterprise customers willing to pay for support and reliability. That's your briefing for Saturday, September 12th, 2026. For DX Today, stay curious.