AI Disclosure

1. AI Systems We Use

MicHelper uses a specialized third-party AI inference provider. AI providers are selected based on performance, security, and data protection requirements. MicHelper does not develop, train, or fine-tune any proprietary or custom AI models.

1.1 Speech-to-Text (STT)

Audio recordings are sent through OpenRouter to an approved AI transcription provider. Audio is processed in real time and is not retained by approved AI inference providers after transcription is complete.

1.2 Chunk Analysis

Individual transcript segments (chunks) are analyzed against customer-defined sales scripts and rules using approved AI inference providers routed through OpenRouter. The AI generates:

• Script compliance scores per conversation segment
• Violation flags and improvement suggestions
• Performance metrics per chunk

1.3 Global Reasoning

After chunk-level analysis, an approved AI inference provider performs reasoning across the entire conversation to produce:

• Overall compliance scores and performance ratings
• Summary analytics and trend analysis
• Consolidated recommendations

1.4 AI Assistant

MicHelper provides an AI-assisted chat powered by an approved AI inference provider routed through OpenRouter, for answering user questions about the platform and sales methodology.

1.5 Technical model stack — confidential

The specific AI models, model sizes, providers, fallbacks, and the routing between them constitute confidential technical information and are not disclosed publicly. OpenRouter routes requests only to approved inference providers that satisfy MicHelper's privacy and Zero Data Retention requirements. The technical model stack may change without affecting the stated privacy safeguards.

2. AI Data Handling

2.1 Zero Data Retention (ZDR)

MicHelper routes AI inference through OpenRouter under Zero Data Retention (ZDR) configurations. Under ZDR, approved AI inference providers do not retain API inputs or outputs for model training or any other purpose beyond the immediate processing request. No customer data sent to the AI inference providers is stored, logged, or used for model improvement.

2.2 Data Flow

The AI processing pipeline works as follows:

1. Audio is uploaded to MicHelper servers
2. Audio is sent to OpenRouter for transcription via encrypted connection (TLS 1.2+)
3. Transcript is returned and stored on MicHelper servers (encrypted at rest with AES-256-GCM)
4. Original audio file is permanently deleted immediately after successful transcription
5. Transcript chunks are sent to OpenRouter for per-chunk analysis
6. Chunk results are aggregated and sent to OpenRouter for global reasoning
7. Final analysis results (scores, flags, recommendations) are stored; transcripts persist during the subscription period for analytics and coaching purposes

2.3 Encryption

All data sent to OpenRouter and approved AI inference providers is encrypted in transit using TLS 1.2 or higher.

3. AI Limitations & Disclaimers

AI analysis is provided for informational purposes only. Accuracy is not guaranteed. You acknowledge and agree that:

• Transcription may contain errors, especially with background noise, accents, or technical terminology
• Compliance scores and violation flags are algorithmic assessments, not definitive judgments
• AI analysis does not constitute legal advice, official compliance certification, or evidence of employee misconduct
• All AI-generated outputs should be reviewed by a qualified human before being used as a basis for employment decisions, disciplinary actions, or compliance determinations
• MicHelper is not liable for decisions made based on AI-generated outputs without adequate human review

4. EU AI Act Preparation

Preparation — high-risk classification contemplated

This page is informational and reflects preparation for a possible future EU market launch. It does not represent current EU AI Act conformity, certification or registration.

4.1 Classification analysis

Under the EU Artificial Intelligence Act (Regulation (EU) 2024/1689), MicHelper's AI-powered performance monitoring system would likely be classified as a high-risk AI system if placed on the EU market, under:

• Annex III, point 4(b): AI systems intended to be used to make or materially influence decisions on the terms of work-related relationships, conditions of work, including task allocation and monitoring or evaluation of persons in such relationships.

This is because MicHelper's AI system monitors and evaluates employee performance in workplace settings, generating compliance scores and performance metrics that may influence employment-related decisions.

4.2 MicHelper's role

If and when MicHelper is launched in the EU, MicHelper would be classified as a Provider (not merely a deployer) of a high-risk AI system. As the developer of the AI-powered monitoring and analysis system, MicHelper would bear the primary regulatory obligations for the AI system's design, development and conformity.

4.3 Current status (important)

• MicHelper is not currently placed on the EU market.
• No conformity assessment has been completed.
• No CE marking is in place.
• No EU registration has been filed.
• The information about future EU availability is forward-looking.

4.4 Why not "limited risk"?

"Limited risk" classification under the EU AI Act applies to systems such as chatbots and deepfake generators, which require only transparency obligations. Workplace monitoring and employee performance evaluation AI systems are explicitly categorized as high-risk under Annex III, requiring comprehensive compliance obligations beyond mere transparency.

4.5 Penalties (for context)

Non-compliance with the EU AI Act may result in significant penalties for providers placed on the EU market:

5. Provider Obligations (MicHelper) — preparation

If and when MicHelper is launched in the EU, the following obligations under the EU AI Act would apply to MicHelper as a Provider of a high-risk AI system:

• Risk Management System: Maintaining a risk management system throughout the AI system lifecycle
• Data Governance: Ensuring training, validation, and testing data meets quality criteria
• Technical Documentation: Maintaining comprehensive technical documentation of the AI system
• Record-Keeping: Automatic logging of events throughout the AI system's lifecycle
• Transparency: Providing clear information to deployers (customers) about the AI system's capabilities and limitations
• Human Oversight: Designing the system to enable effective human oversight by deployers
• Accuracy, Robustness, Cybersecurity: Ensuring appropriate levels of accuracy, robustness, and cybersecurity
• Conformity Assessment: Conducting conformity assessments as required

These obligations are prepared for but not yet fulfilled. No conformity assessment has been completed.

6. Customer Obligations (current — Ukraine)

As the Customer (employer / data controller), you always have the following obligations, regardless of EU AI Act status:

• Inform Employees: Inform employees that AI is used to transcribe and analyze their workplace conversations and generate performance metrics.
• Human Review: Ensure that a qualified human reviews AI-generated outputs before making employment decisions (hiring, firing, promotion, demotion, disciplinary actions).
• Human-in-the-Loop: Maintain meaningful human oversight of the AI system's outputs — do not use AI scores as the sole basis for consequential decisions.
• Inform Workers' Representatives: Where applicable, inform workers' representatives (works councils, unions) and affected workers.
• Data Protection Impact Assessment: Conduct a DPIA where required by applicable data protection law.
• Monitoring: Monitor the AI system's operation and report any serious incidents or malfunctions to MicHelper.

7. No AI Training on Customer Data

MicHelper does not use customer audio recordings, transcripts, or analysis results to train, fine-tune, or improve any AI models. MicHelper does not develop or maintain any proprietary or custom AI models. No fine-tuning is performed on customer data.

All AI inference is routed through OpenRouter under strict Zero Data Retention (ZDR) configurations, meaning approved AI inference providers do not retain input or output data for any purpose beyond the immediate API request.

Aggregated, anonymized usage statistics (e.g., average processing times, error rates) may be used to improve the MicHelper platform itself, but never individual customer content.

8. Contact

For questions about our AI systems or this disclosure:

• Official contact channel: Telegram — @Ooooaala
• Website: michelper.com