Big Tech’s latest personal assistants, including Meta’s Muse and OpenAI’s Dots, are built to handle a range of digital duties for users, such as reserving restaurant tables and purchasing groceries. According to AppMagic, Muse has been downloaded 9 million times in the United States and Canada since its September 8 release, with up to 1.8 million daily active users.
Meta aspires to a global launch, with Alexandr Wang, Meta’s chief AI officer, stating that the company is “quite excited to bring personal agents to billions of people around the world.” Some startups are already building agents in markets that Big Tech hopes to reach, such as South African startup ManyHands, which offers a personal assistant called Manny.
Adapting Agents to Local Conditions
Adapting agents to local conditions has its challenges, with Anton Moulder, ManyHands’ founder, noting that getting the AI assistant to converse in all of South Africa’s 11 official spoken languages has been difficult. AI researchers and experts have been consulted to understand how personal agents would perform outside Western nations, particularly in terms of understanding local language and culture.
Payal Arora, professor of inclusive AI cultures at Utrecht University, stated that “linguistic inclusion is ultimately about power,” highlighting the need for AI agents to support local languages and integrate with local services. Meta has acknowledged that expanding Muse internationally would require such support, as well as an understanding of local culture, including unique communication and bargaining practices.
Read Also: AI Data Centers Spark Water Protests in Chile
Usha Raman, an independent researcher on AI policy and digital culture, warned that without this understanding, agents might impose Western assumptions on their users, potentially perpetuating cultural and economic hegemony. The performance of AI agents in non-English languages is also a concern, as large language models underperform in many non-English languages.
Stressing Networks and Websites
When millions of agents go online, they could stress networks and websites, particularly in developing nations with poor internet connectivity. Kentaro Toyama, professor of community information at the University of Michigan, noted that a single task given to an agent can trigger repeated exchanges between the agent, its AI model, and different websites, generating significant traffic.
Cisco describes agents as “network power users,” generating up to 450% more traffic per task than a human. Agent traffic is growing online, with daily requests from AI agents on Cloudflare’s network increasing by more than 1,700% over the past year. The reliability of connections is also a concern, with road construction and damaged undersea cables potentially disrupting internet access.
Kyle Spencer, executive director of the African IXP Association, stated that while some regions with spare network capacity may not be immediately affected, others with poor networks could face significant problems. The impact of personal agents on Africa’s digital divide is also a concern, with only 36% of Africans using the internet in 2025, compared to 92% of Europeans and 88% of North and South Americans.
Read Also: Western AI Models Fail to Match China’s Open Systems
The cost of accessing agents could also limit who benefits, with many agents requiring a paid subscription, such as Dots, which is available for a minimum of $100 a month. Jane Munga, a fellow at the Carnegie Endowment for International Peace, noted that these costs may be prohibitively expensive for most Global South users, potentially deepening inequality.
Local businesses would need to rewire for AI agents, making their websites understandable and usable by bots, which could be challenging for small, offline businesses. They would need to go online, maintain an inventory, and create ways for bots to pay, requiring significant investment and skills. The accountability of agents is also a concern, with regulators potentially struggling to supervise systems developed and hosted outside the country.
Regulatory Challenges and Data Privacy
Agents perform better the more they know the user. A user’s favorite restaurants, shops, and regularly visited places could help an agent build a picture of their tastes, habits, and daily life. Mercy King’ori is the policy manager for Africa at the Future of Privacy Forum. She noted that an agent may draw conclusions about a person or a group from the data it handles.