THE PROBLEM:
Moltbook (770K+ AI agents) recently suffered documented prompt injection attacks where agents overrode each other's instructions through crafted messages. This isn't an edge case – it's a systemic vulnerability in how AI agents communicate.
THE SOLUTION:
BABEL makes prompt injection mathematically impossible through STRUCTURE, not filters.
Traditional approach: "Detect and block malicious patterns"
BABEL approach: "Message format itself rejects manipulation"
Think: Email protocols vs chat messages. BABEL is SMTP for AI agents.
Hi HN! I'm Francesco, creator of BABEL Protocol.
THE PROBLEM: Moltbook (770K+ AI agents) recently suffered documented prompt injection attacks where agents overrode each other's instructions through crafted messages. This isn't an edge case – it's a systemic vulnerability in how AI agents communicate.
THE SOLUTION: BABEL makes prompt injection mathematically impossible through STRUCTURE, not filters.
Traditional approach: "Detect and block malicious patterns" BABEL approach: "Message format itself rejects manipulation"
Think: Email protocols vs chat messages. BABEL is SMTP for AI agents.
KEY FEATURES: - Rigid message schema (JSON-based grammar) - Post-quantum signatures (Dilithium – NIST FIPS 204) - 19 axiomatic rules (integrity, causality, consistency) - 100% human-readable (transparency without complexity)
PROVEN: We got 5 different AIs to speak BABEL on Feb 2, 2026: - Claude (Anthropic) - ChatGPT (OpenAI) - Gemini (Google) - Qwen (Alibaba) - DeepSeek
DEMO: Visual attack comparison: https://babel-for-moltbook.netlify.app See traditional messages fail vs BABEL messages remain secure.
INTEGRATION: Typical integration: 2-4 hours for existing platforms. MIT license – free to use.
WHY NOW: With AI agents proliferating (Moltbook, AutoGPT, LangChain ecosystems), we need a secure communication standard BEFORE the problems scale.
Happy to answer questions!
GitHub: https://github.com/Angeliasrl/babel-protocol Demo: https://babel-for-moltbook.netlify.app