China’s AI economy had its statement day. DeepSeek is reportedly raising $1.5 billion at a $71 billion valuation ahead of a potential 2027 IPO — a growth story built while US chip restrictions were supposed to prevent exactly this. Alibaba’s Qwen will power Apple Intelligence features in China, per a regulatory filing. And a compression lab showed a 27-billion-parameter model running locally on an iPhone in under 4GB of memory, retaining about 90% of its performance by its maker’s account. The restrictions were the policy; the receipts are the rebuttal.
The watchdog conversation compounded into day two. Coverage of Hassabis’s call now frames it against the government’s crackdown on Anthropic’s Mythos models, and a War on the Rocks analysis argued for an AI threat fusion center before the next such incident. South Korea went the other direction entirely: a government tender for a universal basic AI chatbot and public-services agent, built on local models with state-provided GPUs. One government wants frontier AI checked before release; another wants it issued like a utility.
The courtroom docket grew again. Publishers’ class action against Google over Gemini training on Google Books titles now includes, per one report, an internal document warning of billions in exposure. Meta faces a suit from 26 former employees alleging its AI layoff tools discriminated against workers on disability or family leave. And OpenAI, undeterred by its own legal week, is reportedly betting on a screenless AI speaker as its first hardware act — a proactive home companion, with all the trust questions that phrase implies.
What People Are Asking
What is DeepSeek worth now?
DeepSeek is reportedly in talks to raise $1.5 billion at a $71 billion valuation, with an IPO planned for 2027 — a milestone for China's AI sector despite US chip restrictions. Caproasia →
Is Apple using Alibaba's Qwen?
In China, yes — a regulatory filing shows Alibaba's Qwen will be integrated into Apple Intelligence to power on-device AI features for Apple devices sold there. TechNode →
Why are publishers suing Google?
Major publishers allege Google trained Gemini on copyrighted books from Google Books and Google Play without authorization — with one report citing an internal Google document warning of billions in potential exposure. Eastern Herald →
What is South Korea's universal basic AI chatbot?
South Korea issued a government tender for a universal basic AI chatbot and an AI agent for government services — required to run on local models, with GPUs provided by the state. The Register →
Why is Meta being sued over AI layoffs?
Twenty-six former employees allege Meta's AI systems used in layoff decisions discriminated against workers on disability or family leave. Gizmodo →
Can a serious AI model really run on an iPhone?
PrismML's Bonsai 27B — a compressed 27-billion-parameter model — reportedly runs locally on an iPhone in 3.9GB of memory while retaining roughly 90% of the original model's performance. Gigazine →
Is OpenAI building hardware?
Reportedly yes — a screenless AI speaker positioned as a proactive home companion, facing significant trust and privacy questions before launch. SquaredTech →
What is an AI threat fusion center?
A proposed structure for public-private intelligence sharing on advanced AI risks — argued for in a War on the Rocks analysis as the lesson of the Mythos incident, before the next one. War on the Rocks →
Every Story From July 15
Jump to: Policy and regulation (8) · AI labs and models (7) · AI infrastructure (4) · Consumer tech and culture (4) · Labor and workplace (3) · Society, politics, and ethics (3) · Crypto and fintech (1) · Markets and macro (1) · Other (1)
Policy and regulation8 stories
Google AI chief Demis Hassabis calls for a US-led global watchdog to test powerful AI models before release, inspired by the FINRA model, following a government crackdown on Anthropic.
Major publishers sue Google for using copyrighted books from Google Books and Google Play Store to train its Gemini AI without authorization, citing an internal document warning of billions in potential fines.
The article argues for establishing a structured threat fusion center for AI to improve public-private collaboration and address risks from advanced AI models, using the Mythos incident as a case study.
South Korea has issued a tender for a universal basic AI chatbot and an AI agent for government services, requiring local models and providing GPUs.
The article argues that sovereignty of AI models is not a safety property, using Soofi S as an example to illustrate that ethical alignment can be easily undone and that safety depends on rigorous testing and requirements, not origin.
The Australian Greens criticise the Labor government for failing to legislate enforceable AI protections, and call for a moratorium on new datacentres until rules are in place.
An opinion piece discussing the importance and unique challenges of AI governance for government entities, referencing a new book on the topic.
The article analyzes the legal and policy shift from disinformation to Foreign Information Manipulation and Interference (FIMI) within the EU framework, advocating for a rights-based, multilevel architecture over content control.
AI labs and models7 stories
DeepSeek is raising $1.5 billion at a $71 billion valuation ahead of a potential IPO, highlighting China's AI growth despite US chip restrictions.
PrismML releases Bonsai 27B, a compressed 27B-parameter AI model that runs locally on an iPhone with 3.9GB memory, retaining ~90% of original performance.
Alibaba's Qwen AI model will be integrated into Apple Intelligence, enabling AI features on Apple devices in China as part of a regulatory filing of on-device generative AI services.
China AI startup DeepSeek is reportedly in talks to raise $1.5 billion at a $71 billion valuation and plans an IPO in 2027, following previous funding rounds and growth since its founding in 2023.
Publishers sue Google alleging Gemini AI trained on copyrighted books from Google Books, seeking damages and injunction.
Taskade publishes a market overview of AI app building in 2026, including funding stats and a market map.
A directory-style page listing recent LLM releases from various providers, updated as of July 2026.
AI infrastructure4 stories
Nebius and Meta announce major AI compute deals, refuting overcapacity concerns and highlighting insatiable demand for advanced infrastructure.
OpenClaw's July 2026 release includes major UI, app, model support, and integration updates with numerous fixes and contributions.
n8n's July 2026 release notes list bug fixes across editor, AI Assistant, core, and nodes, improving stability and functionality.
Consumer tech and culture4 stories
An analysis of OpenAI's reported plan to build a screenless AI speaker that aims to be a proactive home companion, highlighting the trust, data privacy, and market challenges it faces based on past smart speaker failures.
The article discusses how AI is helping South African retailers understand changing consumer behavior, with strategies for affordability, loyalty, and personalization.
A news report analyzing FTC data on summer travel fraud, highlighting states with highest losses and report rates.
Reality show contestant Zaid Darbar shares that his wife Gauahar Khan told him her ex Kushal Tandon is a good person, maintaining cordial relations on the show.
Labor and workplace3 stories
Twenty-six Meta employees are suing the company, claiming its AI systems used in layoffs discriminated against those on disability or family leave.
LTM extends Microsoft 365 Copilot to 140,000 L&T Group employees, embedding AI agents into HR, IT, and sales workflows.
Hotels are increasingly using AI to manage operations like housekeeping and maintenance to address labor shortages and high costs.
Society, politics, and ethics3 stories
The article details new, subtle techniques for electoral manipulation in Europe, including reactivating dormant social media accounts, repurposing groups, hijacking hashtags, and using AI-generated content, as discussed at a EU seminar in Sofia.
The article reports a surge in personality rights disputes as AI enables large-scale misuse of celebrities' identities, prompting courts to grant faster and broader relief.
Education experts argue that math instruction must shift from procedural memorization to student-centered, conceptual understanding to prepare students for an AI-driven world.
Crypto and fintech1 story
Saudi Arabia's fintech sector has exceeded its original target of 230 firms, reaching 281, and is driving economic diversification through coordinated policy, regulation, and investment.
Markets and macro1 story
Kevin Ryan's venture firm AlleyCorp closes a $335 million second fund to continue backing early-stage startups in healthcare, deep tech, and general tech.
Other1 story
A music release pack for DJs containing 15 GB of tracks across various genres.
About the RUDI Daily. Responsible Use of Digital Intelligence, daily. Compiled each day from same-day reporting across the web — every story links to its original publisher. How we build it →
Putting AI to work? RUDI helps organizations adopt AI responsibly — training, governance, and hands-on implementation. AI training · Consulting · Talk to us