Will developing nations be able to handle millions of AI agents?

Will developing nations be able to handle millions of AI agents?

Big Tech’s newest personal agents — Meta’s Muse and OpenAI’s Dots — promise to book tables at restaurants, buy groceries, and carry out other types of digital grunt work for users.

Consumers are opting in. Muse has been downloaded 9 million times in the U.S. and Canada since its launch on September 8, market intelligence company AppMagic told Rest of World. As many as 1.8 million people use the app daily.

Meta aspires to a global launch. However, some startups are already building agents in markets that Big Tech hopes to reach. For instance, South African startup ManyHands offers Manny, a personal assistant designed to order groceries, book appointments and make payments through WhatsApp. But adapting agents to local conditions has its challenges. Anton Moulder, ManyHands’ founder, told Rest of World that getting the AI assistant to converse in all of South Africa’s 11 official spoken languages has been difficult.

Rest of World spoke to AI researchers and other experts to understand how these personal agents would perform outside Western nations.

READ THE STORY >

Dispatches from the ground 🌏︎

This week: China proposes export controls on its most powerful AI models, and a Saudi OpenAI engineer is accused of bugging an executive conference room.

China

The Chinese government has proposed export controls on the most powerful AI models. The Ministry of Commerce published a draft regulation, proposing adding a number of AI, chipmaking, biotech, and solar technologies to the country’s export restriction list. AI models trained on the country’s “core data,” such as military and political data, would be banned from exports. Models that are trained on “important data” and are considered powerful — such as those having more than 2 trillion parameters or ranking among the world’s top 10 models — would face export restrictions. The government said the measures align with a global trend to address AI’s threats to national security and human safety, and that it continues to encourage open-source models. But analysts have also interpreted the additions as a way for China to protect its strategic technologies amid U.S. export controls.

— Viola Zhou from New York City

Middle East

A Saudi national working as an engineer at OpenAI was arrested on October 9, accused of planting a hidden recording device inside the company’s executive conference room in San Francisco, the U.S. Department of Justice said. Abdulrahman Alfozan, 35, was charged with using a device to intercept oral communications. The conference room was regularly used for discussions involving proprietary research, product plans, government engagements, and business strategy, the DOJ said. Security cameras showed Alfozan near the room on multiple dates although he had no authority to access it, according to the complaint. A second recording device was allegedly found in a bag he left in another room alongside his wallet and ID. Alfozan, who had worked on OpenAI’s product team since July 2025, had booked a one-way flight to Saudi Arabia for October 12, three days after his arrest. He was fired by OpenAI and faces up to five years in prison if convicted.

— Indranil Ghosh from Abu Dhabi

Stat of the day

— That is how far top Chinese open-weight AI models trailed their leading U.S. rivals on LiveBench benchmark scores. Chinese models can perform some tasks 50% to 90% more cheaply than leading American frontier models, helping attract startups and foreign governments. U.S. restrictions on advanced chip exports have pushed Chinese developers to build systems that use less computing power. “The Chinese companies have been doing a very good job of figuring out how to do more with less,” DigiChina editor-in-chief Graham Webster said at a recent Rest of World event.

You’re invited: Does AI really change the world?

This Friday, join Rest of World for a conversation with Kentaro Toyama, a professor at the University of Michigan and the author of Geek Heresy. The book looks at how AI is upending our assumptions about tech and social change. We’ll explore his famous “Law of Amplification,” the idea that technology merely magnifies human intent, and examine whether persuasive models and autonomous agents are breaking that rule by replacing human capacity altogether. We’ll wrap up with an audience Q&A.

Friday, October 23, 9 a.m. ET / 2 p.m. U.K. / 6:30 p.m. IST.
Virtual event

One more thing

The World Bank is urging developing countries to invest in specialized AI models that run on basic smartphones with limited connectivity, The Guardian reported. That emphasis on local utility mirrors an argument Mozilla CTO Raffi Krikorian made to Rest of World: While consumer hype centers on closed frontier chatbots, organizations are increasingly turning to open models that run on local infrastructures, cut costs, and can be customized to their own languages and date needs.

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

Want to reach a tech-savvy global audience? 🤝 Partner with us.

Like our journalism? Consider donating.