EmbeddingGemma 2: an open, lightweight multimodal embedding model
deepmind.google·6d ago
AI Engineering
Hacker News·Aug 28·5 min read
Show HN: Conduct, open-source guardrails for LLM and MCP tool calls
TL;DR
Conduct is an open-source framework that enforces runtime governance for AI agents, ensuring every action is validated and logged. It provides a secure way to manage LLM calls and tool interactions with cryptographic proof.
✦ Why It Matters
Implement Conduct to enhance security and governance in your AI applications today.
Key Takeaways
- Implement Conduct to enforce AI action policies before execution.
- Utilize Discovery mode for initial visibility into AI actions.
- Integrate with various LLM providers using Conduct Router.
- Leverage pre-built compliance packs for regulatory adherence.
How It Works
Conduct operates by intercepting every action taken by AI agents through its Guard and Router components. The Guard engine evaluates each action against a set of signed policies, determining the appropriate response (block, warn, audit, or inject) before execution.
This proactive approach contrasts with traditional runtime firewalls that only log actions after they occur. The system's integrity is maintained through a hash-chained audit log, ensuring that any tampering is easily detectable.