Skip to content
This is NOT an official site of the Government of Canada. Click here for the official AI registry.

Helpful Places

Government of Canada – DTPR AI Register

Back to your search

AI-Assisted Audio Transcription and Translation for Investigations

Enforcement · Translation & Language Access

What it collects that can identify you

Biometric
Identifiable data
  • Lawfully obtained audio recordings captured during RCMP investigations. These recordings contain voice prints and spoken content from individuals who may be subjects, witnesses, or other parties involved in an investigation.
Run by
Royal Canadian Mounted Police (RCMP)
Where
No fixed location
Kept
Not stated by the Helpful Places.
Shared with
Accountable organization

What it is for

EnforcementTranslation & Language Access

This system uses artificial intelligence to transcribe and translate audio recordings lawfully collected during RCMP investigations. It is used exclusively by Government of Canada employees and is built on OpenAI's open-source Whisper speech recognition model. The audio processed by this system contains personal information, and individuals should be aware that their recorded speech may be transcribed automatically by AI.

What it collects and what happens to it

Data taken in

Biometric
Identifiable data
  • Lawfully obtained audio recordings captured during RCMP investigations. These recordings contain voice prints and spoken content from individuals who may be subjects, witnesses, or other parties involved in an investigation.

Processing

Speech & Audio
  • The core processing technique is automatic speech recognition using OpenAI's open-source Whisper model. Whisper performs both speech-to-text transcription and cross-language translation of spoken audio.

What it does

Sensing (Perceptive AI)
Human executes
  • The Whisper model takes in raw audio recordings and produces structured transcriptions. The AI senses spoken language and converts it to text; investigators then use the text output in their work.
Understanding (Semantic AI)
Human decides
  • For the translation capability, the system understands the meaning of spoken content across languages and produces a translated text output. Human investigators decide how to act on the resulting transcripts and translations.

Outputs

Generated content
Identifiable data
  • The system produces transcriptions (text renderings of spoken audio) and translations (rendered in a target language). These outputs contain personal information derived from the voice recordings of individuals involved in investigations.

Run by

Royal Canadian Mounted Police (RCMP)
  • The RCMP developed and deployed the Voice2Text application, leveraging the open-source OpenAI Whisper model, to assist investigators with audio transcription and translation. It is the accountable federal institution for this AI system.

Voice2Text application — AI Register entry

Built by

OpenAI
  • OpenAI developed and released the Whisper automatic speech recognition model as open-source software. The RCMP's Voice2Text application is built directly on Whisper; OpenAI is listed as the vendor in the AI register entry.

Voice2Text application — AI Register entry

Kept for

Not stated by the Helpful Places.

Shared with

Available to the accountable organization
  • Outputs are available to RCMP investigators and other Government of Canada employees who are authorized users of the system. The register indicates primary users are GC employees.

Stored

Not stated by the Helpful Places.

How to read the colours

Can it identify you?

Anonymized data
Data about people with the link to who is broken. Stripped of identifiers, blurred, aggregated, or noised so this system can’t reasonably tie a record back to an individual.
Pseudonymous data
Each person’s data is tied to a token (hash, ID, template) that lets this system recognise the same person across events, but the token itself doesn’t reveal a name. Reidentification is possible with extra information.
Identifiable data
The data either contains a direct identifier (name, address, account name, recognisable face or voice, plate number) or carries a token this system uses to look up legal identity during processing.

Who completes the loop?

Human decides
This mode suggests; a person decides what to do next. The AI is always advisory — a human is in the loop on every decision. Example: a triage tool ranks cases for a clinician who chooses which to see first.
Human executes
This mode decides; a person carries out the result. Example: an optimizer plans the day’s trash-collection routes, and drivers run them.
Autonomous
This mode decides and acts on its own. No person reviews each decision or carries out the resulting action.

Definitions from the DTPR standard. Amber is about your data, violet about who decides. The fuller the shape and the deeper the colour, the more identifying the data or the less a person is involved.

What you can do

Ask about this system

Questions go to the Helpful Places, not the vendor.

Your rights

  • Right to Be Informed of AI UseThe AI register entry confirms that AI use is disclosed to users of this system. Individuals whose audio is processed may have rights to be informed under the Privacy Act (Canada) and applicable RCMP policies. Contact the RCMP for more information.
  • Right to AccessIndividuals may have the right to request access to personal information held by the RCMP, including transcriptions of their recorded audio, under Canada's Privacy Act. Submit an access to information request to the RCMP's Access to Information and Privacy (ATIP) Office.

Risks and safeguards

  • Civil liberties harmAI-generated transcriptions of investigative audio recordings involving personal communications may affect individuals' privacy and, if used as evidence, their right to a fair process. Transcription errors could misrepresent what was said.Safeguard: Audio is lawfully obtained, use is limited to GC employees (investigators), and AI use is disclosed. Human investigators review outputs before any consequential use; the AI assists rather than replaces human judgment.
  • Reputational harmTranscription errors or translation inaccuracies — inherent in automated speech recognition, especially with accents, dialects, or noisy audio — could misrepresent statements made by individuals, potentially affecting how they are perceived in an investigation.Safeguard: The system is designed to assist investigators, who retain responsibility for verifying and interpreting outputs. The Whisper model is well-documented; its known limitations should inform how outputs are used.