🚀 China’s Open AI Is Setting the New Standard

New data from Stanford HAI shows a big shift: China has taken the lead in open-weight models. In 2025, Chinese institutions pulled ahead of U.S. peers, with Alibaba’s Qwen models even overtaking Meta. Let’s dive in!

In today’s insights:

LATEST DEVELOPMENTS

Evolving AI: Stanford data shows China now leads global adoption of open-weight AI models.

A new Stanford HAI analysis shows Chinese institutions overtook U.S. peers in open-weight AI during 2025. Alibaba’s Qwen surpassed Meta’s Llama as the most downloaded model family, and most new fine-tuned models now build on Chinese bases. Export controls pushed Chinese labs to focus on efficient architectures and permissive licenses, narrowing performance gaps with top closed systems.

Read the report (PDF) here →

Open-weight AI is starting to look less like a sideshow and more like the default “build layer” for a lot of teams, and China is increasingly setting that layer. Stanford HAI’s read is that Chinese model families like Qwen are now driving a big share of downloads and fine-tunes, which means your next app, agent, or internal copilot may inherit Chinese defaults in tooling, prompts, and licensing without anyone making a big strategic decision. That’s great news for cost and access, especially if you want to run models on your own infra instead of paying API tolls, yet it raises procurement questions fast: security testing, jailbreak resistance, and content risk need to be part of vendor review, since U.S. evaluations found DeepSeek models far more susceptible to common jailbreak techniques and more likely to echo misleading narratives.

AI gets smarter when your input is complete. Wispr Flow helps you think out loud and capture full context by voice, then turns that speech into a clean, structured prompt you can paste into ChatGPT, Claude, or any assistant. No more chopping up thoughts into typed paragraphs. Preserve constraints, examples, edge cases, and tone by speaking them once. The result is faster iteration, more precise outputs, and less time re-prompting. Try Wispr Flow for AI or see a 30-second demo.

Start flowing free

Evolving AI: Google removed AI Overviews from some medical queries after reports of misleading health info.

After the Guardian flagged inaccurate medical summaries in Google Search, the company removed AI Overviews from queries like normal liver blood test ranges. The concern was that generic numbers ignored age, sex, or background, risking false reassurance. Some query variations briefly still showed AI content, though results now often point to reporting on the issue. Google says clinicians reviewed the cases and that many sources were reliable, framing the change as part of wider improvements to health search.

Google pulling AI Overviews from a couple of liver-test searches is a small change with a loud signal: when a query looks like something people might use to judge their own lab results, Google is willing to hit the off switch fast. The catch is that AI answers can still pop up through query phrasing and AI Mode prompts, so the risk shifts from one bad summary to lots of near-duplicates that need constant cleanup. For clinicians and health orgs, it raises the bar on publishing clear context in plain language, since medical queries are a heavy trigger area for AI summaries. This is another reminder that AI summaries will expand, then get selectively limited in sensitive corners, reshaping trust and traffic right at the moment someone is anxious and searching.

Evolving AI: OpenAI acqui-hires the team behind Convogo to boost AI cloud work.

OpenAI is bringing in the three founders of Convogo, an AI platform built for executive coaches and HR teams. The deal focuses on the people, not the product, and Convogo will wind down. The startup grew from a weekend project into a tool used by thousands of coaches and major leadership firms. The founders say their real insight was learning how to turn new model releases into useful professional workflows.

This kind of deal is telling you something real about where the AI world is right now: talent is the new battleground. Big players like OpenAI aren’t just buying products anymore, they’re scooping up entire small teams because experience in turning AI into usable tools is rare and valuable. We’ve seen huge moves like the billion-plus dollar Statsig deal, where OpenAI built internal chops for product testing and rollout control rather than just raw models. That pattern shows what’s shaping up across tech: the companies that win will be the ones that can combine smart people with disciplined product practices and ship useful features fast, not just impress with research demos.

Deel's free 2026 trends report cuts through all the hype and lays out in plain terms what HR teams can really expect in 2026. You’ll learn about the shifts happening now, the skill gaps you can't ignore, and resilience strategies that aren't just buzzwords. Plus you’ll get a practical toolkit that helps you implement it all without another costly and time-consuming transformation project.

Get the free report today.

QUICK HITS

🤝 Meta announces a slew of nuclear energy agreements.

🚫 X could face UK ban over deepfakes, minister says.

📈 Demand for AI, tech experts pushes UK financial sector vacancies up 12%.

 *partner link