China’s Manus AI: A Game-Changer or Just Another Overhyped Agent? 

Source: Shutterstock

A little-known Chinese startup has created a new AI product that has gone viral. Feeling a sense of déjà vu? No, we are not talking about DeepSeek-R1, which caused major panic amongst the Western tech companies. This is a new AI agent named Manus, which is rapidly becoming the talk of the global AI community. 

Named after the famous motto Mens et Manus - meaning "mind and hand"- Manus AI is being billed as a completely autonomous agent. In a YouTube video introducing the agent, Manus AI chief scientist Yichao "Peak" Ji claims the agent is “the next paradigm of human-machine collaboration” and offers “a glimpse into AGI”. 

“Manus operates as a multi-agent system powered by several distinct models,” shared Ji. “It works asynchronously in the cloud, which means you can close your laptop anytime, and Manus will notify you when everything is complete.” 

Butterfly Effect, the startup behind Manus AI, claims that as an AI multipurpose agent, Manus can be used for more than traditional AI tasks. It serves as an interface for controlling multiple models and autonomously completing complex workflows, such as generating reports, conducting research, and analyzing financial data. 

(Source: Shutterstock)

Manus AI is currently available in private beta, with access granted through invitation codes only. This means not everyone will have the chance to try it firsthand. Not unless the company decides to expand access or open it to a wider audience. 

"The current invite-only mechanism is due to genuinely limited server capacity at this stage," Zhang Tao, Manus AI's product partner, explained in a social media post. "The current version of Manus is still in its infancy, far from what we aim to deliver in our final product."

One major factor behind Manus AI's growing popularity is its impressive performance in benchmarks. Reports suggest it outperformed OpenAI's o3-powered Deep Research agent and other top-tier models, according to data showcased on the Manus website. The results on the GAIA benchmark, a comprehensive AI test developed by Meta AI, Hugging Face, and the AutoGPT team, were particularly impressive. 

These claims, backed by practical demonstrations, have convinced early adopters and AI enthusiasts that Manus might be among the most effective autonomous agents currently available.

Sounds like a second DeepSeek moment. Maybe, but not quite as groundbreaking.

This ability to control a user’s computer or its programs is not unique or new. OpenAI’s Operator agent and Anthropic’s Computer Use mode are just two of many AI models that offer such capabilities. The Manus AI team may argue that what sets their agent apart is its ability to integrate these functions into a fully autonomous system. 

The benchmark results are impressive, but due to server shortages and the invitation-only access, only a limited number of people have been able to test Manus AI. As more users gain access, we’ll get a clearer picture of its real-world performance, potential limitations, and whether it truly lives up to the hype. 

Source: Shutterstock

Manus is also generating some skepticism about whether the agent represents genuine technological innovation or simply builds on existing large language models (LLMs). It may be a well-optimized orchestration of existing AI tools rather than a true leap forward in AI. 

DeepSeek has to contend with a lot of scrutiny related to data protection and security, and a lot of that was based on where it is located. It appears Manus AI is trying to avoid that. The company has its team based in China, but it also maintained a legal entity in Singapore. However, questions remain about where its servers are located and whether there are any corporate/political affiliations or data transfers to China. 

There is a lack of transparency, and it extends to its safety. The creators of Manus AI claim that the agent can handle tasks without needing humans involved, but what safety measures are in place? Without proper human oversight or proper guadrails, any errors or mistakes by the agent could be disastrous. 

Researchers from Hugging Face published a paper warning that fully automated AI agents should not be developed as they pose security risks, ethical issues, and loss of human control. 

Whether Manus AI is a technological breakthrough or another overhyped AI product, it is evident that Chinese AI products are capturing the imagination of the global AI community. The rise of Chinese AI products indicates that the U.S. has stiff competition in the race for AI supremacy, and Western labs will need to step up their efforts to keep up.