Startups recalibrate Chinese AI for the West
Rest of World/iStock
The concept of machine unlearning — or removing data from an AI system after it has been trained — was largely an academic pursuit when Ben Luria, an Israeli entrepreneur, co-founded his startup. Then, China’s DeepSeek launched its first open-weight large language model last year, and Luria immediately saw the business potential of realigning it to Western values.
A slew of Chinese open-weight models from DeepSeek, Moonshot, Z.AI, Alibaba, and Tencent have won over users — including American businesses and startups — with their lower costs, strong capabilities, and customization options. Others fear the models are too aligned with the Chinese Communist Party, and that they are not as safe as closed systems.
“What we're trying to do is find a way to de-risk the adoption of open-weight models,”Luria, founder of AI safety startup Hirundo, told Rest of World. Luria’s team edits the weights of open models to harden them and make them more secure against cyberattacks, increase their fairness, and reduce their dangerous knowledge of, for example, bio weapons, he said.
Dispatches from the ground 🌏︎
This week, China’s first fully AI-generated film draws a very small audience, and the UAE builds a massive AI research hub.
China
China’s first fully AI-generated feature-length film is in cinemas, but few want to watch it. The sci-fi film, Sanxingdui: Future Past, tells the story of how ancient artifacts provide solutions to a global crisis triggered by AI superintelligence. It features AI-generated characters, scenes, and dialogue, yet took more than 100 people and two years to make. It grossed only 1.45 million yuan ($217,000) in the first week after its October 23 release, far short of its $1.5 million budget. Despite tens of thousands of screenings nationwide, it now averages fewer than one viewer per screening, according to movie data site Maoyan. AI-generated microdramas designed for smartphones have hooked millions of Chinese viewers. But moviegoers still want to see human acting when they spend 100 minutes sitting in a theater.
— Viola Zhou from New York City
Middle East
The United Arab Emirates is building a dedicated AI research facility almost the size of two football fields in Dubai, expanding a campus inside the Dubai International Financial Centre that already hosts more than 500 AI companies. The UAE has spent tens of billions of dollars over the past two years buying chips, building data centers, and investing in U.S. AI companies as it works to build an economy that does not depend on oil. The new building in the financial center’s Zabeel District will provide space for AI research, development, and collaboration between technology companies, government bodies, and investors.
The expansion comes as major foreign AI firms set up in the UAE. Scale AI, a $29 billion U.S. company that tests whether AI systems are accurate and safe enough to be trusted with important decisions for OpenAI, Meta, and the U.S. government, announced at the Dubai AI Festival on October 26 that it had opened a permanent operation inside the campus. Scale has also agreed to work with Presight, an Abu Dhabi AI company that builds data analysis tools for UAE government agencies, on AI systems for government customers.
— Indranil Ghosh from Abu Dhabi
Stat of the Day
— The share of Europe-Asia internet traffic funneled through undersea cables under the Red Sea. Almost all of it squeezes through a choke point less than 32-kilometer (20-mile) wide off Yemen, where regional conflict leaves severed lines facing months of repair delays. To keep the global internet moving in case of rupture, cloud giants like Google, Meta, and Microsoft are locking down expensive overland fiber routes across Iraq and Turkey as alternatives.
And one more thing...
More than 400 deceptive apps posing as everyday AI tools secretly operated as hubs for generating violent deepfakes after slipping past review teams at Apple, Google, and Meta, Wired reported. The Hong Kong-registered network’s reliance on thousands of paid social campaigns lays bare a pattern our reporting examined, where Meta's platforms repeatedly serve as a porous launchpad for offshore bad actors exploiting weak verification safeguards.
Thanks for reading! Send us your feedback at [email protected]. If you enjoyed this newsletter, please forward it to a friend — Edited by Tolu Olasoji
Powered by beehiiv