AI Data Use Policy

1. Core Commitment: No Training on Customer Content

Secure Stream does not use Customer Content to train artificial intelligence or machine learning models. This is a foundational commitment that applies across all Services, features, and workflows.

Customer Content is processed solely to provide, deliver, and support the Services requested by Customer. It is not used for model development, benchmarking, training datasets, or any purpose beyond delivering the Services.

Secure Stream does not use Customer Content to train AI or machine learning models. Where Secure Stream uses third-party providers to process Customer Content, Secure Stream uses commercially reasonable efforts to contractually restrict those providers from using Customer Content for model training except as expressly disclosed or authorized by Customer.

3. Deterministic Processing vs. AI-Enabled Processing

Secure Stream uses two distinct categories of automated processing:

Deterministic Processing. Deterministic processing uses structured calculations, configured rules, mathematical formulas, and predefined logic to generate outputs. The results are reproducible and predictable given the same inputs and configuration. JuryDNA analytics and core Audio Xtract processing are examples of deterministic processing.

AI-Enabled Processing. AI-enabled processing uses machine learning models, statistical classifiers, neural networks, or large language models to generate outputs. Results may vary and are probabilistic rather than deterministic. Transcription, diarization, captioning, optional vendor sentiment, and optional summarization features are examples of AI-enabled processing.

This distinction matters for Customers in regulated industries, litigation contexts, and research environments where the nature of the processing methodology may affect the admissibility, reliability, or defensibility of outputs.

4. Third-Party AI Providers

The Services use third-party AI and machine learning providers for certain features, including transcription, captioning, speech recognition, diarization, and optional vendor-based analytics. These providers are listed in the Secure Stream Subprocessor List.

Contractual Restrictions. Secure Stream uses commercially reasonable efforts to maintain contractual terms with providers that restrict the provider from using Customer Content for model training except as expressly disclosed or authorized by Customer, prohibit the provider from sharing Customer Content with third parties except as necessary to provide the service, require the provider to maintain appropriate security measures, and require the provider to delete Customer Content after processing is complete or within the agreed retention period.

Enforcement Limitations. While Secure Stream uses commercially reasonable efforts to require these restrictions, Customers should be aware that Secure Stream's ability to audit or enforce third-party compliance is limited to the contractual and oversight mechanisms available to Secure Stream. Secure Stream will promptly notify affected Customers if it becomes aware that a subprocessor has materially violated its data use obligations.

5. Data Flow and Processing

When Customer Content is processed through AI-enabled features, the following general data flow applies:

• Customer Content (audio, video, text, or metadata) is transmitted from Secure Stream's infrastructure to the applicable third-party provider via encrypted connections.
• The provider processes the content and returns the output (transcript, caption, diarization data, sentiment indicators or vendor sentiment outputs where enabled, or other results) to Secure Stream.
• Secure Stream stores the output as part of Customer Content and makes it available through the Services.
• The original content sent to the provider is deleted by the provider after processing, subject to the provider's contractual retention obligations.

6. Output Accuracy and Limitations

AI-enabled outputs (transcripts, captions, diarization, optional vendor sentiment outputs, and summaries) are generated through probabilistic processing and may contain errors, inaccuracies, omissions, or misattributions. Deterministic outputs are generated from available inputs, configured rules, and mathematical calculations, but their usefulness and accuracy may still depend on the quality and completeness of the underlying data.

The accuracy and quality of AI-enabled outputs depend on input quality, including audio clarity, microphone placement, speaker overlap, background noise, accents, dialects, source file quality, and network conditions.

Secure Stream outputs are not official records, legal advice, expert opinions, professional conclusions, or substitutes for Customer's independent review and professional judgment. Customers are responsible for reviewing all outputs before relying on them for legal, business, research, compliance, or other purposes.

7. Customer Controls and Opt-Out

Customers have the following controls over AI-enabled and optional automated processing:

• Optional features (such as optional vendor sentiment, summarization, and advanced analytics) may be enabled or disabled by Customer through the Services, where supported.
• Core features (such as transcription, captioning, and diarization) rely on AI-enabled processing and cannot be disabled without affecting the applicable Service.
• Customers who require processing restrictions beyond what is described in this policy should contact Secure Stream to discuss available options and any applicable limitations.
• Customers may request deletion of Customer Content at any time in accordance with the Data Retention Policy.

8. Compliance and Regulatory Considerations

Secure Stream designs its AI data use practices to support Customers in regulated industries, including legal, research, and compliance-sensitive environments. Key compliance considerations include:

• No training on Customer Content, supporting confidentiality and privilege protections.
• Deterministic analytics for JuryDNA, Audio Xtract, and Secure Stream sentiment indicators, supporting reproducibility and defensibility.
• Contractual restrictions on subprocessors, supporting data protection and privacy compliance.
• Transparency about which features use AI-enabled vs. deterministic processing, supporting Customer due diligence and disclosure obligations.

9. Changes to This Policy

Secure Stream may update this AI Data Use Policy from time to time to reflect changes in the technologies used, the Services offered, or applicable law and industry standards. Material changes will be communicated by email, through the Services, or by posting an updated version with a revised effective date. We will provide at least thirty (30) days' notice before material changes take effect.