Hong Kong's Answer to Using OpenClaw Safely

Why It Matters

By providing a verifiable, permission‑based framework for AI agents, Clonet could unlock the convenience of autonomous assistants while safeguarding user data, accelerating mainstream adoption and influencing global AI governance standards.

Key Takeaways

• Clonet creates a secure, identity‑verified AI agent network.
• Agents collaborate across users without exposing full personal data.
• Role‑based permissions limit data access and prevent overreach.
• Human owners intervene when agents exceed predefined boundaries.
• Traceable actions improve accountability for AI‑driven transactions globally.

Summary

Hong Kong’s Generative AI Research and Development Center unveiled Clonet, an open‑source network designed to make AI assistants like OpenCloud safer by assigning each agent a verified digital identity and strict data‑access rules. The platform addresses a core weakness of current agents, which operate in isolated silos and require unrestricted access to users’ accounts, raising privacy and security concerns.

Clonet’s architecture introduces role‑based permissions and a “digital passport” that lets agents recognize, trust, and cooperate with one another across different users without exposing private information. Each agent is limited to the data it is authorized to share, and any request that exceeds its boundaries triggers a human‑in‑the‑loop approval, ensuring accountability and preventing unauthorized transactions.

The video illustrates a holiday‑planning scenario where two users, Sarah and a friend, task their respective agents to compare flights and budgets. The agents negotiate directly, only exchanging data they are permitted to share, and pause for human confirmation when costs exceed the set limit. All actions are logged, allowing users to trace decisions such as a mistaken flight booking.

If adopted widely, Clonet could pave the way for broader consumer acceptance of autonomous AI agents by mitigating security risks, offering regulatory bodies a model for oversight, and encouraging developers to embed collaborative, privacy‑preserving features into future AI services.

Original Description

For more on this story: https://sc.mp/17b43f

A Hong Kong-based AI research body is working on an innovative solution to OpenClaw’s security risks – through a network that allows AI agents to talk to one another, while keeping their actions on the record and within predetermined boundaries. Dubbed ClawNet, this new initiative from the Hong Kong Generative AI Research and Development Centre, or HKGAI, could revolutionise our relationship with AI agents. SCMP’s Eunice Xu is here to explain how.

Subscribe to our YouTube channel for free here: https://sc.mp/subscribe-youtube

Follow us on:

Website: https://scmp.com

SCMP Knowledge: https://www.scmp.com/knowledge

Facebook: https://facebook.com/scmp

Twitter: https://twitter.com/scmpnews

Instagram: https://instagram.com/scmpnews

Linkedin: https://www.linkedin.com/company/south-china-morning-post/