Therapy withVR

EU AI Act Compliance Statement

Version

1.0.0

Effective date

30 days after notice is sent to existing users

Company

withVR BV

Registered address

Jozef Hebbelynckstraat 21, Merelbeke 9820, Belgium

VAT / company number

BE-0790.909.294

Governing law

Belgian law

Jurisdiction

Courts of Ghent, Belgium

Contact

hello@withvr.app | legal@withvr.app

This document sets out withVR BV's assessment of Therapy withVR's obligations and status under the EU AI Act (Regulation (EU) 2024/1689). It is intended for Data Protection Officers, legal and compliance teams, procurement officers, and institutional users who require documentation of the platform's AI governance position.

This statement covers: the AI features present in Therapy withVR; the risk classification assessment; applicable obligations and their current status; and the obligations that do not apply and why.

1. AI Features in Therapy withVR

Therapy withVR includes the following AI-powered features:

Feature

Detail

Avatar voices

All avatar speech inside VR is synthesized using Google Text-to-Speech. Always active in all sessions. The person inside VR hears a synthetic voice, not a human recording.

Sentence translation (optional)

Translates conversation text between languages. Uses OpenAI API. Off by default, user-activated.

Text generation (optional)

Generates suggested conversation sentences from a topic or prompt. Uses OpenAI API. Off by default, user-activated.

Autocorrect (optional)

Corrects spelling and grammar in text entered by the supervising professional. Uses OpenAI API. Off by default, user-activated.

Whisper speech recognition (optional)

Converts spoken words to text during session setup. Uses OpenAI API. Off by default, user-activated.

Speaker grammar (optional)

Adjusts gendered grammar to match the avatar's voice. Uses OpenAI API. Off by default, user-activated.

Formality adjustment (optional)

Adjusts formality level of AI-generated text. Uses OpenAI API. Off by default, user-activated.

Emotional speech (optional)

Adjusts avatar voice characteristics to match the avatar's set emotion. Uses OpenAI API. Off by default, user-activated.

Important: All optional features are off by default and must be deliberately activated by the supervising professional. The platform generates scenario content to support communication practice. It does not evaluate, score, diagnose, or make decisions about any person's health, learning, or eligibility for services.

2. Risk Classification Assessment

2.1 Overall classification: Not a high-risk AI system

Therapy withVR has been assessed against the EU AI Act's risk classification framework and is not classified as a high-risk AI system under Article 6 and Annex III of the Regulation.

Classification: Therapy withVR falls within the limited risk category under the EU AI Act, with specific transparency obligations under Article 50 applicable from August 2, 2026. It is not a high-risk AI system. No conformity assessment, EU database registration, or CE marking is required.

2.2 Annex III assessment

Annex III of the EU AI Act lists the categories of AI use cases that are classified as high-risk. The following table sets out each relevant Annex III category and the assessment for Therapy withVR:

Annex III category

Assessment for Therapy withVR

Education and vocational training (Annex III, point 3) - AI that evaluates learning outcomes, monitors and detects prohibited behavior, or determines access to educational institutions

Not applicable. Therapy withVR does not evaluate learning outcomes, monitor student behavior, or determine educational access or placement. The platform provides a virtual speaking environment. All educational and clinical decisions are made by the supervising professional, not by the AI.

Healthcare (Annex III, point 5) - AI intended to be used as medical devices, or that assists clinical decision-making in a manner that could harm individuals

Not applicable. Therapy withVR is not a medical device (not CE marked, not FDA regulated). It does not assist clinical decision-making, score assessments, analyze health data, perform patient triage, or influence treatment decisions. The platform creates practice scenarios. Clinical decisions remain with the supervising professional.

Employment and workers management (Annex III, point 4)

Not applicable. Therapy withVR is not used for hiring, performance evaluation, or employment management.

Law enforcement, migration, justice (Annex III, points 6-8)

Not applicable.

Critical infrastructure (Annex III, point 2)

Not applicable.

Biometric identification (Annex III, point 1)

Not applicable. No biometric identification is performed. Voice synthesis (Google TTS) generates audio from text - it does not identify or authenticate individuals.

2.3 Article 6(3) exemptions (belt-and-suspenders)

Even if any AI feature in Therapy withVR were considered to touch an Annex III category, the Article 6(3) exemptions would apply. The AI features in Therapy withVR:

2.4 Prohibited practices assessment (Article 5)

Article 5 of the EU AI Act (in force since February 2025) prohibits certain AI practices outright. The following have been reviewed for Therapy withVR:

Prohibited practice

Assessment

Subliminal or manipulative techniques that distort behavior

Not present. Avatar voices and scenario content are fully visible and controlled by the supervising professional.

Exploitation of vulnerabilities related to age, disability, or socio-economic status

Not present. The platform supports communication practice. It does not exploit any vulnerability. Users with communication differences interact with it as a practice tool under professional supervision.

Social scoring systems

Not present. No scoring, ranking, or profiling of individuals based on social behavior is performed.

Real-time remote biometric identification in public spaces

Not present. No biometric identification is performed.

Emotion recognition in employment or education contexts

Not present. The platform does not recognize, analyze, or infer emotions from users or participants.

AI-generated or manipulated content used to deceive

Not applicable in the platform's context. Avatar voices are disclosed as AI-synthesized (Article 50 disclosure).

Conclusion: No prohibited AI practices under Article 5 have been identified in Therapy withVR.

3. Applicable Obligations and Current Status

The following EU AI Act obligations apply to Therapy withVR:

Obligation

Article

Status

Detail

AI literacy

Article 4

In force Feb 2025 - Compliant

withVR BV provides live onboarding training to all users covering AI feature usage, limitations, and responsible use. AI feature documentation is included in platform documentation available at withvr.app.

Prohibited practices review

Article 5

In force Feb 2025 - Compliant

Review completed. No prohibited practices identified. See Section 2.4.

Transparency for AI-synthesized voices and AI-generated content

Article 50

From Aug 2, 2026 - Preparing

Disclosure is already present in platform documentation, Privacy Policy, EULA, and Terms of Service. In-platform disclosure will be confirmed compliant before August 2, 2026.

4. Obligations That Do Not Apply

The following EU AI Act obligations apply only to high-risk AI systems and do not apply to Therapy withVR:

Obligation

Reason it does not apply

Conformity assessment (Article 43)

Therapy withVR is not a high-risk AI system under Annex III.

Technical documentation under Annex IV

Required only for high-risk AI systems.

Quality management system (Article 17)

Required only for high-risk AI system providers.

Risk management system (Article 9)

Required only for high-risk AI systems.

Data and data governance (Article 10)

Required only for high-risk AI systems that use training data.

Logging and record-keeping (Article 12)

Automatic logging requirements apply only to high-risk AI systems.

Human oversight measures (Article 14)

Article 14 obligations apply to high-risk AI systems. Human oversight is maintained by design as a clinical best practice, but Article 14 does not impose formal requirements on this platform.

Accuracy, robustness, and cybersecurity (Article 15)

Required only for high-risk AI systems.

EU database registration (Article 49 / Article 71)

Not required for AI systems outside Annex III scope.

CE marking

Not required. Not a high-risk AI system. Not a medical device.

5. Article 50 - Transparency Obligations

Article 50 of the EU AI Act imposes transparency obligations on providers and deployers of certain AI systems. These obligations apply from August 2, 2026. The following Article 50 obligations apply to Therapy withVR:

5.1 AI-synthesized voices (Article 50(4))

Article 50(4) requires that persons exposed to AI-generated audio be informed that the content is artificially generated or manipulated. In Therapy withVR, all avatar voices are synthesized using Google Text-to-Speech. The person inside VR hears a synthetic voice in every session.

Disclosure is addressed as follows:

5.2 AI-generated text content (Article 50(4))

When optional OpenAI features are activated, some session content (sentences, prompts) may be AI-generated. The supervising professional sees and controls all content before it is presented in VR. Content that originates from AI features is presented to the supervising professional to review, modify, or discard before use.

Disclosure is addressed in the same documents listed in Section 5.1.

5.3 Deployer obligations for institutions

Under Article 50, deployers (institutional users) have their own transparency obligations toward end users. Institutions using Therapy withVR should ensure that:

6. Third-Party AI Providers and Data

Provider

Data sent

Google Text-to-Speech

Text strings for voice synthesis only. No personal data about users, clients, or participants in normal operation.

OpenAI API

Text entered by the supervising professional into optional AI-powered fields (when features are activated). User-entered text only - not client names, session recordings, or participant PII. Under OpenAI's API data usage policy, API inputs are not used to train OpenAI's models by default.

No client names, session recordings, or personally identifiable information about participants are sent to OpenAI or Google as part of normal platform operation. Supervising professionals are advised not to enter participant identifiers into any AI-powered text field.

7. Information for Deploying Institutions

Institutions deploying Therapy withVR in clinical, educational, or research settings should be aware of the following:

Topic

Detail

Classification

Therapy withVR is not a high-risk AI system. Institutions do not have high-risk deployer obligations under Article 26 with respect to this platform.

AI literacy (Article 4)

Institutions are responsible for ensuring that their staff who use Therapy withVR have a sufficient level of AI literacy, as required by Article 4 (in force since February 2025). withVR BV's onboarding training supports this obligation.

Transparency to participants (Article 50)

Institutions are responsible for ensuring that participants are informed of AI involvement where required by professional standards, institutional policy, or applicable law. withVR BV provides disclosure documentation to support this.

Professional oversight

All clinical, educational, and research decisions are the sole responsibility of the supervising professional. AI features in Therapy withVR generate supporting content only and do not replace professional judgment.

DPIA

A Data Protection Impact Assessment (DPIA) is not required for Therapy withVR on AI Act grounds alone, as the platform is not a high-risk AI system. Whether a DPIA is required under GDPR Article 35 for your specific use case is an institutional determination. withVR BV can provide documentation to support a DPIA on request.

8. EU AI Act Key Dates

Date

Obligation

February 2, 2025

Article 5 (prohibited practices) and Article 4 (AI literacy) in force

August 2, 2026

Article 50 (transparency for AI-generated content, synthetic voices) in force. withVR BV's disclosure obligations under Article 50 apply from this date.

August 2, 2026

High-risk AI system obligations under Annex III in force. Not applicable to Therapy withVR.

August 2, 2027

High-risk AI system obligations for AI embedded in regulated products (Annex I) in force. Not applicable to Therapy withVR.

9. Contact and Further Information

For questions about this compliance statement, to request supporting documentation, or to discuss Therapy withVR's AI governance position:

withVR BV

Jozef Hebbelynckstraat 21, Merelbeke 9820, Belgium

General enquiries: hello@withvr.app

Legal and compliance: legal@withvr.app

withVR BV | Jozef Hebbelynckstraat 21, Merelbeke 9820, Belgium | legal@withvr.app | withvr.app