AI and Synthetic Voice Policy

AI Features

Mojo White may use AI systems for OCR, parsing, summarization, assistant responses, podcast generation, and text-to-speech workflows.

Model and provider availability may change as Mojo White improves quality, reliability, privacy, and cost controls.

AI features may use Anthropic, OpenAI, Groq, DeepInfra, Google Cloud, Microsoft Azure Speech, or other providers listed on the Subprocessors page. Some features may fall back to another provider if the primary provider is unavailable.

Outputs may be incomplete, inaccurate, outdated, biased, unsafe for a particular context, or inconsistent across repeated runs. Users must review outputs before relying on, sharing, submitting, or publishing them.

Assistant answers may miss sources, misread documents, invent citations, misstate statistics, or produce confident-sounding analysis that is wrong. Users are responsible for checking source documents and independent references before relying on outputs.

Transparency and Labeling

Generated summaries, assistant responses, podcast scripts, and synthesized audio are AI-generated or AI-assisted outputs.

Mojo White intends these disclosures and generated-output labels to support AI transparency obligations, including synthetic-content notice requirements such as EU AI Act Article 50 where applicable.

Mojo White may label generated audio through product UI and metadata where practical, such as ID3 metadata fields indicating the output was produced by Mojo White and AI-generated.

If you export, publish, distribute, submit, or play generated audio or text outside Mojo White, you are responsible for any disclosure required by law, platform rules, employment/school policies, election rules, advertising rules, or audience expectations.

Do not remove synthetic-audio labels, metadata, watermarks, provenance signals, or warnings that Mojo White or a provider applies unless you have a lawful reason and the removal does not mislead anyone.

No Training and Provider DPAs

Mojo White does not use customer content to train Mojo White or third-party provider models as part of normal product use.

Haro LLC treats dated DPAs or equivalent terms for Anthropic, OpenAI, Google Cloud, Microsoft Azure, Groq, DeepInfra, WorkOS, Stripe, Apple, Cloudflare, Upstash, Firebase/Google Analytics, MaxMind, VirusTotal, and other active providers as vendor-review records.

No-training commitments can depend on account type, endpoint, region, feature, provider settings, abuse-monitoring exceptions, feedback submission, research opt-in, and future provider terms. Haro LLC periodically reviews provider settings and retains configuration evidence where available.

Users should not submit secrets, highly sensitive data, or regulated third-party information to AI features unless they have authority and the use is appropriate for the feature, provider, and jurisdiction.

Safety Boundaries

Mojo White may refuse or limit requests involving self-harm methods, violence, weapons, malware, exploitation of minors, deceptive impersonation, or other prohibited uses.

These limits apply even when the request is framed as research, fiction, or quoted from a user document.

Mojo White may apply automated and human review for abuse prevention, trust and safety, support, legal compliance, provider-policy compliance, and security. Review is limited to what is reasonably necessary for the purpose.

Voice-Specific Rules

• Use voices, cloned voices, personal voices, and voice styles only where you have the necessary rights, consent, and disclosures.
• Do not create synthetic audio that falsely claims to be a real person's statement, consent, testimony, evidence, authorization, emergency message, endorsement, or instruction.
• Do not use generated audio for robocalls, scams, voter suppression, fake customer support, impersonated executives, fake law enforcement, fake medical instructions, or any other deceptive or harmful communication.
• Do not use Mojo White to infer, identify, authenticate, or profile a person from voice data unless a feature expressly supports that use and all required consents and laws are satisfied.

Assistant Boundaries

The AI assistant may answer using your documents, conversation history, and available tools, but it is not guaranteed to cite every relevant source, detect every conflict, or understand legal, medical, financial, technical, or safety context correctly.

Mojo White may limit tool access, context length, uploads, retrieval, web search, or provider calls for privacy, security, policy, cost, or reliability reasons.

Provenance and Audit Follow-Up

Haro LLC may maintain synthetic-audio provenance records, such as ID3 tags, file metadata, generation IDs, provider, model or voice ID where safe to store, timestamp, account ID, and deletion-retention status.

Provenance logs are designed to support abuse investigations and legal requests without storing more content than needed or undermining ordinary deletion rights.