ทันเอไอ

Phinq Launched: AI Agents Security Control System to Block Errors Before Execution

Phinq is a new open-source software acting as an intermediary to filter commands from AI Agents to prevent potential damage.

As AI agents play an increasingly vital role in automation, the risk of AI accidentally executing catastrophic commands—such as deleting critical data or modifying systems unintentionally—also rises. Recently, an open-source project named "Phinq" was developed specifically to solve this problem.

Phinq acts as a Runtime Governance Layer, positioning itself as a proxy or middleware to intercept all tool calls made by AI agents in real time. The system classifies risk levels before allowing commands to actually execute.

In terms of operation, Phinq utilizes a Human-in-the-Loop concept, splitting control into two main parts. First, low-risk or reversible commands are allowed to proceed automatically. Second, high-risk and irreversible commands, such as dropping a database, are temporarily suspended until human approval is granted.

Additionally, the system features a Tamper-evident Audit Log that records decision histories and actions at every step using hash-chained technology to prevent tampering and support post-audit reviews. For compatibility, Phinq supports integration with a wide variety of leading AI frameworks and tools such as OpenAI, Anthropic, LangChain, CrewAI, AutoGen, and Claude Code.