Letheia launches Truth-Safe API for AI safety compliance

Letheia on October 6 unveiled the Truth-Safe API, a developer-facing safety and compliance layer aimed at conversational AI, AI companions and other human-to-system apps. The company says the product is built to help developers prove safety decisions to regulators with auditable records, risk scoring and policy enforcement.

Why it matters: - State lawmakers are adding specific AI safety requirements around risk detection, referrals and reporting. - Letheia is targeting the gap between what statutes require and what developers can prove their systems did. - The pitch is simple: a compliance layer that can show regulators evidence instead of relying on self-reporting or custom internal logs.

What happened: - Letheia officially unveiled the Truth-Safe™ API on October 6 in Stamford, Connecticut. - The product is a developer-first compliance and safety architecture for conversational agents, AI companions and other human-to-system applications. - The API is available now, with developer documentation, API access and pricing information available from Letheia.

The details: - Truth-Safe is a RESTful Safety Orchestration system for applications that accept open-ended human language. - The API sits between human input and system response to detect defined risks, apply safety policy, determine required actions and create an auditable record. - The system provides real-time risk scoring, country-resolved crisis referral information where required, and a record of each safety decision. - Letheia says the product is based on patent-pending innovations. - The workflow is framed as Detect. Decide. Act. Prove. - The API records what was detected, which ruleset version was used, what decision was made and what action the application was told to take. - The system accepts the severity and timing of recent messages alongside the current message, so context can influence how a message is handled. - The API is designed to plug into RESTful microservices and LLM gateway middleware with minimal code refactoring. - Safety detection, severity resolution, crisis resources, a published referral indicator and auditable decision records ship with the API. - The calling application controls what users see, while detection is governed by a versioned, clinically approved ruleset. - The API returns resources matched to the user’s country, with the emergency number ordered first at the highest severity. - Tony Scamurra, Letheia founder and CEO, said the system is meant to automatically identify safety signals, determine severity, execute policies and create an auditable record. - Scamurra said developers get the architecture without building it, and regulators get evidence they can verify.

Between the lines: - The product is aimed at an enforceability problem in AI regulation, not just a tooling problem. - Letheia is positioning safety as a governed process that can be measured and audited, which may appeal to regulators that need repeatable evidence. - The company is also signaling that compliance can be built into the system design rather than added as a manual review step. - Letheia’s white paper, Safety Orchestration for Human-to-System Applications, examines the architecture across twelve jurisdictions and includes evidence from real-world use.

What's next: - Developers can use the new API now through Letheia’s documentation and access channels. - Letheia is offering the product alongside a white paper that may help frame adoption for legal, policy and engineering teams. - State-level AI rules will likely determine how quickly tools like Truth-Safe get adopted in regulated applications.

The bottom line: - Letheia is betting that AI safety compliance will need verifiable, machine-generated evidence, not just policy promises.