Our solutions follow AlayaCare's high-security standards, including proper authentication, authorization, and input validation. AI products follow the same user roles and permissions within the AlayaCare platform. Models are securely hosted, and data pipelines are designed with security in mind.

We self-host models on secure internal compute clusters and use proprietary model services only where they satisfy our security, data sovereignty, and PII/PHI compliance requirements.

We use selective, sanitized, and fully anonymized portions of historical client data to train domain-specific models. Client data is fully isolated for inference, testing, and validation. 

Data is securely stored in siloed databases with restricted access and multi-factor authentication.

AlayaCare performs regular security audits and compliance reviews. We also adhere to responsible AI governance policies, including the Voluntary Code of Conduct on the Responsible Development and Management of Advanced Generative AI Systems

In addition to our policies and governance, we stay abreast of bodies of research focused on AI safety by collaborating directly with Montreal Institute for Learning Algorithms (Mila), who work closely with Law Zero ensuring that AI is safe for humanity. We build comprehensive monitoring, observability and oversight systems for all of our AI-based products and run a comprehensive suite of internal evaluations and benchmarks to measure model performance and associated risks.

All generally available solutions have automated monitoring and observability and are subject to continuous review to ensure accuracy, compliance, and ethical standards.