Our new ASPI report 'The Party’s AI' is out now. It shows how China’s LLMs, vision models and “AI+” governance architecture are hard-coding censorship and control into the future of AI. https://t.co/lOuzKtBchu

https://t.co/4lF4NYKpp0

Huge thanks to my brilliant co-authors: Bethany Allen, Shelly Shih, Stephan Robin, Nathan Attrill, Jared Alpert, Astrid Young & Tilla Hoja. Thanks to the Human Rights Foundation for supporting this work. If you think this research matters, RT to get it in front of more people.

The rest of the report expands the picture: • AI in policing, courts & prisons • AI-driven online censorship • Minority-language LLMs used for surveillance of Uyghur, Tibetan communities • AI-enabled fishing platforms affecting other countries' economic rights

Chinese models are rapidly gaining global users through open-weight releases & easy API access. As I told The Washington Post, if we don’t understand how they’re built, we risk importing censorship & political control hidden inside the technology itself. https://t.co/qlLKWThF7n

In some cases, the same image generated wildly different answers depending on language: • English → neutral description • Simplified Chinese → state-aligned warnings (“cults”, “extremism”) • Traditional Chinese → less constrained

We tested image understanding in English, Simplified Chinese, and Traditional Chinese. We found: • Much stricter censorship in Chinese • Strongest distortions on Tiananmen, Tibet, Falun Gong & Xinjiang • Some models quietly added pro-state framing even without refusing.

The results are staggering. Models routinely: ❌ refuse to answer ❌ erase key details ❌ repeat state narratives Here’s one example: a major Chinese LLM flat-out refuses to describe the 1989 Tiananmen Square massacre.

Chapter 1 tests what this system actually produces. We asked leading Chinese LLMs to describe politically sensitive images: Tiananmen, Hong Kong 2019, Uyghur and Tibetan protests, Falun Gong, Taiwan, and more.

China’s national AI-safety standard categorises 31 types of “unsafe” content — from terrorism and privacy breaches to… 🟥 criticism of the CCP 🟥 “Western ideology” 🟥 peaceful protest This is the logic shaping every model released in China.

The report opens by explaining something crucial: In China, “AI safety” doesn’t mean protecting people from harm. It means ensuring AI protects the state, aligns with “core socialist values”, and avoids anything that might “harm the national image”.

‘Western politicians have failed to speak up about the slaughter of civilians in Sudan because the United Arab Emirates has bought and paid for their silence, according to a top Sudanese general.‘ https://t.co/IuGLRWabRD

‘Fan, who now spends most of her time outside China‘ Where is Fan Bingbing based? https://t.co/RsZpWF7xXY

👀 https://t.co/uCbr86L5or

I just chipped in to help Cairo Takeaway fight a defamation case born of a tabloid hit job. They could use your support too. https://t.co/oTu6qHsBWa

Pretty strong denial from Rudd. https://t.co/J0GvXH2tct

It seems like all the attached photos are missing from the Epstein email dump? Am I getting that right? https://t.co/iw5VRWChLc

I'm starting to hit maximum schadenfreude on this story.

Hedge funds were already trimming their "Magnificent Seven" stakes in Q3 & now Peter Thiel has dumped his entire Nvidia position, just days after SoftBank's big sell-down. Seems significant. https://t.co/UovLOLc8Bd

👀 https://t.co/8qsKBmdNkr