AI trust scoring company Tumeryk Inc. announced a strategic integration with AI encryption firm DataKrypto Co. to launch a joint service that they claim offers the world’s first encrypted guardrails for operational AI security. The new Encrypted Guardrails for Operational Security combines DataKrypto’s real-time encryption of retrieval-augmented generation data, model weights and prompt payloads with Tumeryk’s AI Trust Score, self-calibrating prompt security and responsible AI controls to provide end-to-end protection across the entire AI pipeline. The integrated solution encrypts all stages of generative AI workflows, from data retrieval through model inference and response generation, while simultaneously enforcing compliance and policy alignment. The two companies argue that while traditional AI guardrails focus on monitoring model outputs, they often leave critical data flows vulnerable to attack or misuse. The integration between Tumeryk and DataKrypto closes the gap by encrypting every component of the AI pipeline, from vector embeddings and foundation models to tool-calling prompts and guardrail policies. The result is end-to-end protection that strengthens the AI attack surface while also preventing threats such as data exfiltration, prompt injection and model manipulation. Core to the new solution is DataKrypto’s FHEnom technology, which allows encrypted computation on embeddings and model weights while maintaining hardware-enforced isolation within secure enclaves. Tumeryk complements this with real-time prompt inspection, using self-calibrating guardrails to detect and block noncompliant or potentially harmful inputs before they reach the model.