China’s New Autonomous AI Agent Manus Could Be the Next Deepseek Moment

Chinese AI lab Monica AI has introduced Manus, an agentic AI system capable of operating without human oversight, marking a shift toward artificial intelligence that makes autonomous decisions.

Unlike ChatGPT and Google Gemini, which require human prompts, Manus is designed to function independently, making real-time decisions without oversight. The launch on March 6, 2025, reinforces the ongoing shift towards AI automation, placing it at the center of an international debate on governance, security, and competition.

Manus AI: China’s Fully Autonomous AI Agent Sparks Global Debate on AI Regulation and Security

Manus is designed to execute tasks independently, making real-time decisions without external validation. 

Manus AI distinguishes itself through its so-called multi-signature (multisig) control system, allowing it to independently manage actions across different AI models.

Reports suggest that it employs LLM chaining and reinforcement learning for continuous optimization, enabling it to refine its decision-making processes over time.

Manus AI says its system has already demonstrated autonomous capabilities in tasks such as resume screening, workflow automation, and candidate evaluation.

Benchmark results shared my Manus AI indicate that it has achieved a new state-of-the-art performance level in the GAIA evaluation framework, which assesses AI in reasoning, automation, and tool use.

The GAIA (General AI Assistants) benchmark is a comprehensive AI performance test developed by Meta AI, Hugging Face, and AutoGPT to evaluate AI systems’ real-world capabilities. It assesses an AI’s ability to reason logically, process multi-modal inputs, use external tools effectively, and automate real-world tasks.

GAIA is highly respected in the AI community because it tests an AI’s real-world utility rather than just theoretical knowledge. It reveals a significant performance gap between human and AI capabilities, with human respondents achieving 92% accuracy while top AI models struggle to reach similar levels.

Manus AI’s reported GAIA scores indicate strong performance in complex reasoning and automation, putting it ahead of some of its key competitors, including OpenAI Deep Research, Google’s Project Mariner, and AI agents powered by Anthropic’s Claude models.

How Manus AI Differs from Other AI Assistants

Existing AI models operate under the human-in-the-loop principle, where users approve AI-generated actions before execution.

OpenAI’s Operator AI and Google’s Project Mariner follow this model, requiring explicit human confirmation before carrying out commands. Manus AI breaks from this approach, eliminating the need for direct approval and allowing AI-driven decision-making to occur autonomously.

The system is designed to handle a broad spectrum of tasks, from managing software environments to navigating digital workflows without oversight. This independence could make Manus AI a sought-after tool, with high demand for access.

Following its launch, reports of users emerged about them staying up all night attempting to obtain invite codes, underscoring public interest in the AI-driven automation tool.

According to YiCai Global, Zhang Tao, a partner at Manus, addressed the situation on WeChat Moments, admitting that the company had “completely underestimated the level of enthusiasm” from the public. He clarified that Manus “has never opened any paid channels for invitation codes”, nor has it “allocated any marketing budget” for promotion.

He also emphasized that the company’s priority is to “ensure a stable experience for existing users”, with plans to gradually expand access in a structured manner.

The official website does not provide details about the team behind Manus. However, records from Tianyancha indicate that Butterfly Effect (Hong Kong) Limited, the company behind Monica AI and Manus, was established in 2023 and holds full control over Beijing Red Butterfly Technology Co., Ltd.

The latter was founded by Xiao Hong in 2022 and was responsible for launching Monica, an AI-powered browser extension targeting overseas users. The current legal representative of Beijing Red Butterfly Technology is Feng Qionghua, while company filings show that Xiao Hong stepped down as a shareholder in March 2023. Additionally, Xiao Hong was linked to Wuhan Nightingale Technology Co., Ltd., which withdrew its stake in the company in December 2024.

Security Risks and Regulatory Scrutiny

The ability of Manus AI to self-initiate actions is raising cybersecurity concerns. Experts warn that autonomous AI agents could be misused for cyberattacks, large-scale fraud, or automated disinformation campaigns if deployed without proper safeguards.

The lack of human verification makes it difficult to prevent manipulation, introducing new ethical and legal challenges.

Regulators have been responding with increased scrutiny after the impactful emergence of DeepSeek AI from China as a challenger of OpenAI, Google DeepMind and other leading US AI labs.

Meanwhile, the U.S. government is considering whether fully autonomous AI should be classified as a security-sensitive technology, which could lead to new regulatory frameworks similar to those applied to foreign-built telecommunications infrastructure.

China’s AI Expansion and Strategic Self-Sufficiency

Manus AI’s introduction comes as China intensifies efforts to reduce reliance on U.S. technology and accelerate AI-driven automation. With tightening U.S. AI chip export restrictions limiting access to high-performance semiconductors, Chinese firms are optimizing AI software for greater efficiency.

Besides DeepSeek’s R1 reasoning model, Alibaba’s QwQ-32B is an example of this shift, designed to handle complex reasoning tasks with lower computational demands.

Unlike Western AI firms, which focus on enhancing human-guided AI systems, China’s leading AI startup are investing in fully independent AI models. Manus AI represents a major step in this strategy, demonstrating how AI can function autonomously without cloud-based dependency on U.S. tech firms.

AI Governance and the Global Response to Manus

The release of Manus AI has intensified discussions about AI governance, particularly as no existing regulations explicitly address the risks of fully autonomous AI.

In the U.S., concerns over AI security have already led to legislative discussions on whether Chinese AI models like DeepSeek should be banned from government networks. These debates are now expanding to include fully independent AI agents, with proposals to classify them as high-risk technologies. If adopted, such measures could impose strict licensing requirements or outright bans on AI models that operate without human validation.

Meanwhile, the European Union is exploring regulatory updates to address the rise of autonomous AI. If the EU determines that AI models like Manus lack sufficient oversight, they may introduce new AI accountability laws that limit the deployment of self-governing systems.

China’s Strategic AI Shift and Its Impact

China’s move toward autonomous AI reflects a broader strategy to strengthen its domestic AI capabilities and reduce reliance on Western technologies. With ongoing semiconductor restrictions preventing Chinese firms from accessing high-performance AI chips, developers are focusing on software optimizations instead.

By emphasizing autonomous AI, China can possibly sidestep some of the hardware limitations imposed by export bans.

This approach contrasts with Western AI firms, which continue to prioritize human-in-the-loop AI models. While OpenAI’s Operator ensures that AI-driven actions are approved by users, Manus represents a shift toward full independence. The question now is whether this model will be widely adopted or whether regulatory concerns will limit its global reach.

Manus AI’s introduction could mark another turning point in the AI industry, as it forces developers, businesses, and regulators to reconsider how AI should function in real-world environments.

The balance between AI regulation and technological advancement is now more critical than ever. If self-governing AI is allowed to operate freely, industries that rely on automation could see unprecedented efficiency gains. However, the risks of AI-driven decision-making without accountability remain a key concern.

As AI competition intensifies, the future of autonomous AI will depend on whether companies and governments can establish clear guidelines for its use. Whether Manus AI becomes a widely adopted model or faces restrictions, its arrival is another sign of the beginning of a new era—one in which AI is no longer just a tool but an independent entity capable of making its own choices.