Compliance

The recording never becomes someone else's problem.

Every cloud dictation or transcription tool asks you to accept that a vendor holds a recording of your client, your patient, or your matter. CamoVoice does not send the audio anywhere, so there is no vendor, no retention window, and no contracts nor policies to periodically review.

01 — Why audio and video are problematic

A recording is worse than a document.

Audio and video cannot be effectively redacted before transmission: a recording contains the identity in the media, in the background, and hidden metadata. The moment it is uploaded for transcription or conversion, it's disclosed.

CamoVoice removes the upload. Bundled local models transcribe on the user's device; no service to assess.

No BAA needed

We never create, receive, maintain or transmit PHI. A desktop application that does not contact its vendor is not a business associate.

No processor

Nothing to add to your Article 30 record, no DPA to sign, no sub-processors to monitor for this product.

No retention

Audio is processed in local memory. Transcripts exist where you save them and nowhere else. There is no vendor-side copy to subpoena, breach, or forget to delete.

Not even for licensing. There is no activation call at first launch and no periodic re-check. CamoVoice installs, activates and runs permanently on a machine that has never been connected to a network. Start your packet capture before installation and it will stay empty — a test most offline-marketed software fails.

Who hears the audio: cloud dictation versus CamoVoice In cloud dictation the audio reaches a transcription vendor, its cloud storage and potentially its sub-processors. With CamoVoice the audio reaches only the local model on the user's device. CLOUD DICTATION — PARTIES WITH ACCESS TO THE RECORDING Youspeak STT vendorreceives full audio Cloud storageretention window Sub-processorsregister to monitor Four parties. One BAA, one DPA, one transfer assessment, one breach surface per party. CAMOVOICE — PARTIES WITH ACCESS TO THE RECORDING Your device — microphone → bundled model → text in-memory · no telemetry · no account · no upload
Fig. 1 — The vendor-assessment burden is proportional to the number of parties who hear the audio.

02 — Health information

HIPAA: the business associate question, answered by removing the associate

Duty → mechanism
ObligationWhat it requiresHow CamoVoice discharges it
Business associate agreement45 CFR § 164.502(e); § 164.308(b) A BAA with any party creating, receiving, maintaining or transmitting PHI on a covered entity's behalf. Cloud transcription vendors are squarely business associates. Not engaged. Audio and transcripts never reach us. There is no BAA to execute because there is no associate.
Technical safeguards§ 164.312 Access control, audit controls, integrity and transmission security for ePHI. Transmission security is satisfied trivially: there is no transmission. Access control and encryption at rest fall to the endpoint controls you already run.
Psychotherapy notes§ 164.508(a)(2) Authorisation for most disclosures; notes must be kept separate from the record. Dictate process notes directly to a local file. No third party ever holds the session content, which is what makes cloud dictation untenable for many professionals.
Minimum necessary§ 164.502(b) Limiting disclosure to the minimum necessary. Zero is the minimum. Where you subsequently want AI assistance on the transcript, run it through CamoText first.
Breach notification§§ 164.400–414 Notification following breach of unsecured PHI, including breaches at business associates. We cannot be the source of one. Vendor breach notifications have been a significant share of reported HIPAA incidents; removing the vendor removes that exposure entirely.

03 — European Union

GDPR: voice data, special categories, and the transfer question

Duty → mechanism
ObligationWhat it requiresHow CamoVoice discharges it
Security of processingArt. 32 Measures appropriate to the risk, considering the nature and scope of processing. Keeping the recording on the endpoint is the strongest available measure for audio, because audio cannot be minimised before capture the way text can be redacted before sending.
Processor contractArt. 28(3) Written terms with any processor, plus authorisation and oversight of sub-processors. Not engaged. No DPA is required from us and none is offered, because we process nothing on your behalf.
Special category dataArts. 9, 4(14) Heightened conditions for health data and for biometric data processed for unique identification. Clinical and legal dictation routinely contains Art. 9 health data in its content. CamoVoice performs no speaker identification and creates no voiceprint, so it does not itself generate biometric data under Art. 4(14), and the content never leaves the device regardless.
TransfersChapter V A valid mechanism before personal data leaves the EEA. Not engaged. Most consumer dictation apps route audio to US infrastructure; this one routes it to your own RAM.
Storage limitationArt. 5(1)(e) Retention no longer than necessary. Nothing is retained by default. Files exist only where you choose to save them, under your existing retention schedule.

04 — United Kingdom

SRA confidentiality and attendance notes

Duty → mechanism
ObligationWhat it requiresHow CamoVoice discharges it
ConfidentialitySRA Code of Conduct, para. 6.3Client information kept confidential.Dictations and transcriptions never pass through a third party. The disclosure that would need justifying does not occur.
Legal professional privilegePreservation of privilege over communications and, for litigation privilege, working material.No third-party recipient, so no argument about whether onward disclosure to a vendor affects the privileged status of the note.
UK GDPR / DPA 2018Per the EU analysis above.Same position: no processor, no transfer, no retention by us.

06 — Germany

§ 203 StGB and the works council

Duty → mechanism
ObligationWhat it requiresHow CamoVoice discharges it
Criminal secrecy§ 203 StGBCriminal liability for unauthorised disclosure of secrets entrusted to lawyers, doctors and psychotherapists. Cloud dictation ordinarily relies on the assisting-person route under § 203(3), with its selection and obligation requirements.No disclosure occurs at all, so the assisting-person analysis is unnecessary. For Ärzte and Psychotherapeuten dictating patient records, this is the material difference.
Professional duty§ 43a(2) BRAOComprehensive Verschwiegenheitspflicht.Dictated mandate notes remain within the Kanzlei's own systems.
Works council co-determination§ 87(1) No. 6 BetrVGCo-determination for technical equipment suited to monitoring employee conduct or performance: a provision that stalls most voice-tool rollouts.CamoVoice has no admin console, no usage analytics, no central logging and no telemetry. It cannot report on an employee because it reports nothing to anyone. The REC indicator is visible to the user on their own screen, and recording is manual and explicit.

Practical note for German rollouts. The absence of any central management surface is a deployment trade-off: it also means no fleet-wide policy enforcement. We think that is the right trade for this product, and we would rather state it than let procurement discover it later.

07 — France

Secret professionnel

Duty → mechanism
ObligationWhat it requiresHow CamoVoice discharges it
Secret professionnelArt. 226-13 Code pénalCriminally sanctioned secrecy binding avocats, médecins and other professionals, described as général, absolu et illimité.A dictated note that never leaves the practitioner's device is never revealed to anyone. This is the only posture fully consistent with an absolute duty.
CNIL expectationsMinimisation and documented security measures; particular scrutiny of voice data.Local-only processing, no retention, no telemetry: all independently verifiable.

08 — The obligation we cannot remove for you

Recording consent is still yours to obtain.

Offline processing solves the disclosure problem. It does not solve the capture problem. If you record another person speaking, consent law applies regardless of where the audio is processed.

WhereThe rule, in short
US — all-party statesCalifornia, Florida, Illinois, Pennsylvania, Washington and others require all parties' consent to record a confidential communication. California's is at Penal Code § 632.
US — federal and one-party states18 U.S.C. § 2511 permits recording with one party's consent, but state law governs where stricter.
EU / UKRecording a person creates personal data and needs a lawful basis, notice, and often consent. Employee recording raises separate employment-law issues.
Germany§ 201 StGB makes recording the non-public spoken word without consent an offence, which is independent of data protection law.

Why we point this out. Most dictation is a practitioner speaking alone, where none of this applies. But we would rather flag the one obligation our architecture does not touch than let it be discovered after the fact. CamoVoice shows a visible REC indicator, records only when you press the key, and never listens in the background. This helps evidence that capture was deliberate and bounded.

09 — Scope of these claims

What this page does not say

We do not claimBecause
That CamoVoice makes you compliantIt removes one exposure. Consent, retention, access control on the device, and supervision remain yours.
That transcription is accurate enough to go unreadAccents, cross-talk and specialist vocabulary produce errors, even though Custom Words can help. Anything entering a clinical or client record must be read before it is filed.
That we hold SOC 2 or ISO 27001We do not. Those attest to how a vendor handles your data; we handle none, and it's directly verifiable.
That macOS permissions are automatic Global voice typing requires Accessibility and Input Monitoring access. See the macOS Permissions for Global Hotkey in the User Guide.

Last reviewed July 2026