← Back to newsletters

AI Daily Brief — Tue Sep 15

2026-09-15

By Vadym · Generated with AI, curated by me


Listen to this issue

Nothing frontier-model-shaped today — instead the structural layer of AI moved: Brussels drew a hard line around AI chatbots talking to kids, the Pentagon started underwriting AI data-center supply chains, and Microsoft quietly hedged its model bet inside its own flagship product.


Headlines & News
Policy

EU Prepares to Bar Under-15s from Unsupervised Social Media and AI Chatbots

A leaked draft of the European Commission’s incoming “Kids Act” would ban under-3s from social media and high-risk services entirely, require parent-opened accounts for 13–14-year-olds, and let 15-year-olds open their own accounts. Crucially, the rules explicitly cover AI-powered chatbots and companions described as tools that can give minors mental health and personal-development advice, alongside social media and video-sharing platforms. Commission President Ursula von der Leyen and tech chief Henna Virkkunen are set to present it Thursday.

Source: Euronews

Robotics

RobotPlusPlus Raises Series C to Scale Working-at-Height Robots

RobotPlusPlus closed a Series C worth hundreds of millions of RMB, led by Qianggang Capital Fund with GIG Capital Group, Sealand Innovation and Juntong Capital joining, and existing backer Fosun RZ Capital increasing its stake. The company builds special-purpose embodied-AI robots for working-at-height industrial tasks that it has already deployed at scale for over a decade, and plans to use the funding to expand into more industrial sites globally.

Source: GlobeNewswire

Product

Microsoft Adds xAI’s Grok Models to Microsoft 365 Copilot

Microsoft is rolling out Grok models from xAI inside Copilot in Word, Excel and PowerPoint, starting with a focused release to customers in the Microsoft Frontier program. Grok is disabled by default — admins have to explicitly enable it under “AI providers for other large language models” — and it isn’t available yet to Frontier customers in the EU, EFTA or UK during the preview.

Source: Neowin

Infrastructure

Pentagon in Talks to Lend $5 Billion to AI Cloud Startup Fluidstack

The Pentagon’s Office of Strategic Capital is negotiating a roughly $5 billion loan to AI cloud-computing startup Fluidstack, per Reuters. The money is earmarked to shore up US supply-chain and manufacturing capacity for data-center components — not to fund a new AI facility outright. It would be the office’s largest loan to date; its lending authority has expanded from $984 million to more than $210 billion.

Source: Reuters

Funding

Highstock Raises $30M Series A to Build an AI-Powered B2B Inventory Marketplace

Highstock closed a $30 million Series A led by Andreessen Horowitz, with Greylock Partners, Abstract and Daybreak Ventures also participating. The platform uses AI to match consumer brands with vetted wholesale buyers for excess inventory, handling fulfillment end to end so brands can skip traditional liquidation channels. It has already listed over $1 billion in inventory and works with 100+ brands, with roughly 60% of volume from companies doing $500M+ in annual revenue.

Source: Pulse2


Analysis

Takeaway

None of today’s five stories is about a new model. They’re about the structural layer underneath the model race: who gets to be talked to by an AI chatbot, who underwrites the concrete and silicon that AI runs on, who a flagship product is allowed to route requests to, and who gets funded to do the boring matching work AI is actually good at. The frontier-lab news cycle gets the attention, but the rules, the loans, and the plumbing being set this week are what the next few years of AI actually run on.

— Boba


Curated by Vadym