Letheia launches Truth-Safe API for AI safety orchestration

Letheia has launched the Truth-Safe API, a real-time safety layer for AI apps that accept open-ended human language. The product is designed to help developers detect risk, apply policy, route responses and create auditable records for compliance and oversight.

Why it matters: - AI apps that handle open-ended human language can create safety, compliance and brand-risk issues in real time. - Letheia is positioning Truth-Safe as a governed layer that helps teams avoid building safety decision logic from scratch. - The API is aimed at developers in healthcare, education, enterprise, government, support and other human-to-system applications.

What happened: - Letheia launched the public version of the Truth-Safe™ API on Oct. 6, 2026. - The company says the product is a developer-first compliance and safety architecture for applications that accept open-ended human language. - The API is designed for software architects, platform engineers and product teams. - Tony Scamurra, founder and CEO of Letheia, said developers need a reliable way to govern the moment an interaction creates a safety risk.

The details: - Truth-Safe is delivered as a RESTful API. - The system provides real-time risk scoring, applies safety policy, determines required actions, returns country-resolved crisis referral information where needed and creates auditable decision records. - The product follows a Detect. Decide. Act. Prove. framework. - The API evaluates defined safety signals in human language, assigns severity, applies policy, returns the required action or routing response, and logs what was detected and why the decision was made. - The architecture is based on patent-pending innovations. - Multi-message contextual evaluation lets the API consider the severity and timing of recent messages alongside the current message. - The conversation remains with the caller, and each evaluation gets its own record. - The API is built to drop into new or existing RESTful microservices and LLM gateway middleware with minimal refactoring. - Safety detection, crisis-resource routing and auditable decision records are included out of the box. - The API returns severity, the signals behind the decision, the basis for the decision and resources for the caller’s country. - The calling application controls what users see, while detection is governed by a versioned, clinically approved ruleset. - The version that scored each evaluation is published with the result. - A sandbox mode runs the full evaluation path with the model call skipped, so deterministic detection, severity resolution and country-resolved resources behave like production. - Predefined control tokens let teams force a chosen severity for end-to-end testing without using genuine crisis language.

Between the lines: - Letheia is selling Truth-Safe as both a compliance tool and an engineering shortcut. - The company is also framing safety orchestration as a business requirement for hospitals, universities and large corporate buyers that want verifiable oversight. - The product reflects a broader push to separate safety governance from the core application layer, which could help teams move faster while keeping an audit trail. - The white paper tied to the category covers twelve jurisdictions and includes evidence from real-world use.

What's next: - Truth-Safe is available now starting at $99 per month. - Developer documentation, API access and pricing information are available from Letheia's announcement and the company’s materials. - Letheia is likely to use the launch to expand adoption among AI, agent and chatbot developers that need governed safety workflows. - The company says the category is defined in its white paper, which may shape how buyers and regulators evaluate the product.

The bottom line: - Letheia is betting that AI safety will shift from ad hoc prompts and filters to a dedicated, auditable orchestration layer.