Dr Fazal Ali
The Hundred Model War started in 2023. From Beijing to the southern Silicon Valley of Shenzhen, China is implementing its AI+ project, which was introduced in 2024 and expanded in 2025.
The aim is to accelerate the integration of AI across economic and social sectors by embedding AI in manufacturing, healthcare, consumption, and governance to enhance productivity and stimulate economic growth.
After the release of DeepSeek, the buzz in China was about “OpenClaw”, developed by Peter Steinberger from Austria. OpenClaw sparked a frenzy with people everywhere experimenting with its code to build their own “lobsters”. Powered by open-source data and technology, the code is available to anyone who wants to customise it to develop indigenous AI edge demos.
A big trend is “Bring your Own AI” (BYOAI). But BYOAI may expose proprietary enterprise data in bureaucracies and private companies to unregulated external environments. To address this vulnerability, software provider Kilo launched KiloClaw for Organisations, an enterprise-grade platform built to rein in decentralised agent deployments and restore architectural oversight.
OpenClaw became the fastest-growing AI project in history, amassing 230,000+ GitHub stars in under 14 days. The challenge is that OpenClaw is built for self-hosting. This means you need Docker, a server, and some technical capability to get started. Docker is an open-source platform that enables developers to build, ship, and run applications within isolated, lightweight packages called containers.
It serves as a standardised environment, solving the “it works on my machine” problem by ensuring that software runs identically regardless of where it is deployed, whether on a developer’s laptop or in the cloud.
This is where KiloClaw found an opportunity. KiloClaw is a fully managed hosting service for OpenClaw, launched by Kilo Code, a platform serving 1.5 million+ developers. It takes the powerful OpenClaw Agent and handles all the infrastructure intricacies.
KiloClaw handles: (1) Server provisioning and maintenance, (2) Docker container management, (3) Security configuration, (4) Auto-updates, (5) Process monitoring and auto-restart, and (6) Unified billing through Kilo Gateway.
Using KiloClaw, organisations now have a toolkit to govern Autonomous Agents and discern the presence of “Shadow AI”. Sometimes, companies spend months securing LLMs and formalising vendor agreements, while developers and knowledge workers begin acting independently.
In other cases, workers began bypassing official ICT procurement processes by deploying autonomous agents on their personal infrastructure to automate their daily workflows. This exposed proprietary enterprise data to unregulated external environments.
To address this vulnerability, Kilo launched KiloClaw for organisations. Kilo targets the lack of visibility surrounding personal Agent deployment. When skilled workers set up autonomous agents to parse error logs, or financial analysts in private companies or bureaucracies deploy local scripts to reconcile spreadsheets, they prioritise immediate efficiency over security protocols.
These “Claws” often gain access to corporate Slack channels, Jira boards, and private code repositories through personal API keys. Since these connections occur outside official IT oversight, they may create blind spots for data exfiltration and intellectual property leaks.
KiloClaw provides a centralised control plane for security teams to identify, monitor, and restrict these autonomous actors without hindering their productivity gains.
When OpenClaw exploded in the global tech community, Jensen Huang, the CEO of Nvidia, called OpenClaw the AI era’s answer for ordinary people. Following Huang’s endorsement, Chinese tech giants started releasing apps built on OpenClaw. Secondary school students lined up waiting for free, customised versions. They were curious to find out more about the “lobsters”. The CEO of Cheetah Mobile, Fu Sheng, was astonished at how his lobster easily automated and simplified many routine tasks in his workflow.
Because OpenClaw was built during the 100 Model War, the code remains available to anyone in China who wants to customise it to work with Chinese models. After all, Western models like ChatGPT and Claude are not accessible in China.
However, it was not long before Beijing’s AI security authorities warned developers, workers and students about serious risks associated with improper installation and use of OpenClaw. Some government agencies have since banned workers from installing the tool. The frenzy shifted from offering to install to offering to remove it.
This is part of China’s disorder-and-control top-down method. The hallmark is that control must never signal discouragement. The signal is always for ubiquitous innovation and invention that makes AI security paramount.
The 2025 release of DeepSeek demonstrated the strong entrepreneurial appetite for innovation and research in China despite chip shortages. It also underscored how ordinary citizens are willing to adopt open-source platforms. Once Beijing signals its priorities, demand and supply take over. To drive economic competitiveness, Beijing signals its priorities by subsidising office space, providing seed capital, and incentivising lines of credit. The outcome is secure development built around a constellation of values, beliefs, technologies, working knowledge, and practices that lessen vulnerabilities.
Dr Fazal Ali completed his Master's in Philosophy at the University of the West Indies. He was a Commonwealth Scholar who attended the University of Cambridge, Hughes Hall, the provost of the University of Trinidad and Tobago and the acting president, and chairman of the Teaching Service Commission. He is presently a consultant with the IDB.
