AI Transparency Statement

Mosaic Health AI uses artificial intelligence (AI) to help healthcare communications teams create scientific and medical content more efficiently. This statement explains how AI is used within Mosaic, which models are involved, and what controls ensure safe, compliant, and transparent use.

Mosaic is a SaaS platform that assists medical and creative professionals in generating:

• Structured message platforms (key messages, supporting messages, and linked proof points)
• Reference packs that trace claims to scientific sources

All content is produced within a controlled workflow designed for human review, traceability, and regulatory compliance.

AI supports specific steps of content creation by analysing reference material and producing draft outputs. It is never autonomous — all outputs are reviewed, edited, and approved by qualified professionals before use.

AI is used for:

• Drafting structured text based on uploaded scientific references
• Suggesting message structures or supporting points
• Linking draft content to relevant references

Mosaic’s AI does not make clinical decisions, perform patient assessments, or handle personal health information (PHI).

Mosaic integrates with third-party Large Language Models (LLMs) via secure API connections.

• Primary provider: OpenAI (GPT models)
• Supporting infrastructure: Pinecone for vector database search and retrieval
• All data in transit is encrypted (TLS 1.2+)

No client data is used to train, fine-tune, or otherwise improve any AI model.

• Data Minimisation: Only essential reference content is processed; no patient or personal data is uploaded.
• Data Control: All inputs and outputs remain the property of the client.
• No Training on Client Data: Inputs, prompts, and outputs are never used by model providers for training.
• Data Retention: Files and outputs are deleted within 30 days of pilot completion unless otherwise agreed.
• Encryption: All data encrypted in transit (TLS 1.2+) and at rest (AES-256).

Mosaic follows a strict human-in-the-loop model:

• Every AI-generated output is reviewed by a human user before approval or export.
• Users remain responsible for content accuracy, appropriateness, and compliance.
• AI activity is logged for traceability and audit.

This ensures human judgement, clinical understanding, and brand context always remain in control.

Mosaic’s approach to AI is built on the following values:

1. Accountability – We remain responsible for how AI is used and governed.
2. Transparency – Clients can see when and where AI has been used.
3. Human Oversight – AI supports users; it never replaces expertise.
4. Fairness – AI outputs are designed to reduce bias by focusing on verified scientific sources.
5. Safety and Security – Models and infrastructure operate within GDPR and EU AI Act principles.

Mosaic is classified as a low-risk general-purpose AI application under the EU AI Act. We proactively meet relevant obligations:

• Clear disclosure when AI is used
• Documentation of AI functions and model providers
• Technical and human oversight mechanisms
• No use of biometric, behavioural, or high-risk data processing

We will continue updating this statement as the EU AI Act enforcement phases progress.