Last updated: 21 November 2025
Operated by Healthcare Design Studio Ltd
1. Purpose of This Policy
This AI Policy explains how automated systems are used within TalkFRCA ("the Service") and what users must understand about the limitations, data processing, and rules associated with AI-generated content.
2. How AI Is Used in TalkFRCA
TalkFRCA uses automated systems to:
- Generate exam-style questions, prompts, and viva scenarios
- Deliver simulated examiner responses
- Produce transcripts of sessions
- Analyse session content for performance metrics
- Support future improvements to questioning logic and educational features
Automated systems are used to assist with exam preparation only.
They are not intended for clinical decision-making or patient-facing use.
3. AI Providers Used
TalkFRCA uses selected third-party AI and infrastructure providers, including:
- OpenAI – language model processing and automated generation
- Vapi – real-time interaction orchestration and voice processing
- Deepgram – real-time speech-to-text transcription with medical terminology recognition
- ElevenLabs – voice synthesis
- Supabase – storage and authentication infrastructure (not an AI model provider)
These providers may receive raw transcripts, interaction data, or technical metadata necessary for operation and improvement.
4. Accuracy and Limitations of AI-Generated Content
AI-generated content may contain inaccuracies, omissions, outdated information, or misleading statements.
All AI outputs—including definitions, explanations, physiological descriptions, and viva-style responses—must be independently verified using established educational or professional sources.
Content generated through AI must never be used to guide real-world patient care, clinical decisions, or treatment planning.
5. Data Sent to AI Providers
To operate and improve automated features, the Service may send:
- Raw session transcripts (text only)
- Typed responses
- Session interaction data
- Technical metadata
- Prompt information required for generation
This information may be used by AI providers (OpenAI, Vapi) for model improvement. Deepgram processes audio in real time but does not retain audio data; Vapi may store audio streams and derived metadata on its own infrastructure to monitor quality or train its conversational stack.
TalkFRCA never stores audio locally—only text transcripts are retained in our systems.
6. Voice Processing via Vapi and Deepgram
If you use voice-based interactions:
- Your audio is processed in real time by Vapi (orchestration) and Deepgram (transcription)
- Audio streams are converted to text using Deepgram's medical terminology recognition
- The processed transcript is sent to the Service
- TalkFRCA never stores audio data
- Vapi may store encrypted audio and interaction metadata for a limited retention window to debug quality, prevent abuse, or train its service. Deepgram does not retain audio after transcription.
- Only the resulting text transcript is retained by us
This applies to all present and future voice features; by using them you accept Vapi's handling of audio data.
7. Data Retention for AI-Related Information
AI-related data is retained as follows:
- Session transcripts are stored indefinitely
- After account deletion, transcripts are anonymised and retained
- Interaction logs are held temporarily for processing
- Analytics data may be retained indefinitely in anonymised form
These retention rules are consistent with the TalkFRCA Privacy Policy.
8. User Responsibilities and Restrictions
To ensure safe and lawful use of AI systems, users must not:
- Enter any patient-identifiable information
- Describe real clinical cases involving identifiable individuals
- Submit confidential workplace scenarios or sensitive personal data
- Upload or input restricted, copyrighted, or proprietary exam questions
- Use the Service for clinical decision-making or patient-facing advice
Any violation may result in account suspension.
9. Transparency
TalkFRCA aims to be clear about how AI is used:
- Automated systems generate and analyse educational content
- Interactions may be processed by multiple AI providers
- Some outputs may be manually reviewed to improve quality and safety
- AI systems do not personalise content based on user identity or behavioural profiles
This policy explains how AI operates at a high level without revealing internal logic.
10. Human Oversight
In specific contexts, content or user activity may be manually reviewed by authorised members of Healthcare Design Studio Ltd for the purposes of:
- Quality improvement
- Safety monitoring
- Investigating misuse
- Refining educational frameworks
Manual review is limited to operational and educational needs.
11. Changes to This Policy
We may update this AI Policy from time to time.
Continued use of the Service indicates acceptance of any updated version.