AI and Data Privacy

At VirtualSpeech, we integrate cutting-edge AI features to enhance communication skills training responsibly, transparently, and ethically. This page outlines how our AI tools work, how we handle data, and how we comply with privacy laws such as the European AI Act.

Our Approach to AI

VirtualSpeech uses AI to support learners in practicing presentation, speaking, and interpersonal skills. Our AI-powered tools provide feedback on:

  • Content of your communication (speech or conversation)
  • Delivery of your communication (speech or conversation)

These features are designed to support learning, not to make decisions about learner performance or outcomes. The AI generates objective feedback and scoring based on communication metrics, which are meant solely for personal development.

No AI-Driven Decision-Making

AI-generated scores (e.g., content ratings during roleplays) are intended for training purposes only. They are not used for hiring, promotions, or any other employment-related decisions.

This commitment ensures VirtualSpeech remains compliant with Article 5 of the European AI Act regarding the use of AI in high-stakes contexts.

Body Language Tracking — No AI Involved

VirtualSpeech does not use AI to calculate or interpret body language.

Instead, body language insights (such as eye contact, movement, or posture) are captured using a VR headset and controller sensors only. We do not record or analyze video or camera feeds. No facial recognition or biometric data is processed.

How We Use the OpenAI API

VirtualSpeech integrates the OpenAI Enterprise API to power conversational roleplays and generate AI scores.

  • No data is used to train OpenAI models
  • OpenAI may temporarily retain data for up to 30 days to monitor for abuse
  • We use default API settings with no persistent memory
  • All client data is logically separated and isolated

For more information, see: OpenAI’s Enterprise Privacy Policy

Learner Data and Admin Access

AI feedback (including transcripts and scores) is visible to organizational admins unless the client chooses to anonymize learner data.

If anonymization is enabled:

  • Admins only see cohort-level summaries
  • Individual learners retain full visibility of their own data

This flexibility allows organizations to balance privacy with performance tracking.

Data Storage and Security

All AI-related data is stored on VirtualSpeech’s secure AWS servers (Germany for EU customers, Ohio for others). Data includes:

  • AI-generated scores and feedback
  • Voice recordings (if manually saved by the learner)
  • Roleplay and speech transcripts (deletable by learners or admins)

Access is restricted to senior members of our technical team, solely for debugging or client support. We use industry-standard encryption (SSL/TLS and AES-256) and follow best practices for secure infrastructure.

Transparency and Oversight

VirtualSpeech aligns with the European AI Act through three core principles:

1. Transparency
  • Learners and admins are clearly notified when they’re interacting with an AI system
  • Roleplay scores and voice analysis are visible and explainable
  • No facial, emotional, or biometric data is collected
2. Fairness
  • Learner data can be anonymized at the organization’s request
  • AI is used only to enhance learning, not to judge, rank, or penalize users
3. Human Oversight
  • Admins have access to transcripts and performance data (unless restricted)
  • Final evaluations and outcomes are always determined by human trainers or managers

Questions or Concerns?

If you’re an HR, compliance, or procurement professional and need more detail on how VirtualSpeech handles AI and data privacy, please contact:

Sophie Thompson
Co-Founder, VirtualSpeech
sophie@virtualspeech.co.uk

Or see our general Privacy Policy for full data processing information.