Dev Tools · 1h ago
New Open-Source Tool Cryptographically Watermarks LLM Outputs
resk-mark embeds a verifiable cryptographic watermark into LLM-generated text during token sampling. The watermark survives rewording and truncation, allowing provenance verification with a public key. The open-source Python library hooks into models like Mistral-7B to prove text originated from a specific system.
Meridian48 take
A practical solution to a growing problem, but its robustness against sophisticated adversarial attacks remains to be proven in real-world deployments.
llm-watermarkingcryptographic-provenance