Input data and information sources
Outsiders often also speak, such as telephone customers or hearing participants. Recordings and transcripts contain voices, names and people's statements.
Speech recognition and transcription
Also known as: Automatic speech recognition and transcription
Speech recognition automatically transcribes dictation, conversations and other recordings.
Understand the use case
What defines this use case?
The UK judiciary and Germany's Federal Employment Agency use speech recognition; BaFin reports smaller financial institutions transcribing customer conversations.
This page describes a typical system design. Actual data flows and permissions depend on your implementation.
- System design
- Generative assistance
- Ability to act
- Provides results for subsequent use
- Fields of use
- judiciary, public administration, banking
- Also known as
- speech-to-text, dictation recognition, conversation transcript
How it works, step by step
System design and data flows
Input data
Dictation, a telephone call or another recording arrives as an audio track.
Outsiders often also speak, such as telephone customers or hearing participants. Recordings and transcripts contain voices, names and people's statements.
Processing
Recognition converts speech into words and punctuation; specialist vocabulary supports medical or legal terminology.
Speech recognition processes sections of the audio signal and assigns words to them. These sections form the transcript; depending on the tool, timestamps and speaker identification may be added. The transcript is an interpretation of the recording, rather than directly measured text.
Results
A transcript is available for human review and correction.
A generated result is newly worded or created. To assess it, the instruction and context on which it is based must remain identifiable.
Use and impact
The transcript enters files, findings, decisions or consultation notes.
The model supplies information for subsequent use. Selection, adoption or a professional decision in the following workflow is a separate step and should not automatically be attributed to the model.
Explore the system design: Select a component to see its task, incoming and outgoing data, and role in the use case. The guided walkthrough explains each phase. Then use Yes, Uncertain and No to represent your specific system variant.
8 review first7 clarify8 baseline threats
- Present
- Depends on your system
- Absent
- Review first
- To clarify
How to use the model
The diagram shows the typical architecture. Select a component to see what happens there and adapt its features to your system. Enable “Analysis guidance” to see which threats need reviewing for this architecture. The page's checklist continues to describe the typical case.
Changes from the typical case
None yet. Each change will appear here.
Where most threats arise
- 4 clarify
- 3 review first
- 2 review first
- 1 review first, 1 clarify
In this system design: 8 to review first, 7 to clarify, 8 baseline threats.
Architecture as a list
External access: Depends on your system
To clarify
Input: Part of every AI system
Review first
Baseline for every AI system
Third-party content: Present
AI model: Part of every AI system
Baseline for every AI system
Provider service: Present
Review first
To clarify
Training and feedback: Present
Review first
To clarify
Result: Part of every AI system
Review first
- Factual Inconsistencies (Hallucinations) Review priority
- Model Denial of Service
Baseline for every AI system
person: Part of every AI system
Review first
- Overreliance Review priority
Affected person: Depends on your system
Unapproved AI services: Depends on your system
To clarify
Supply chain and technical foundations: 5 baseline threats applying to every AI system
Baseline for every AI system
How it differs from similar use cases
AI meeting minutes and conversation documentation
Summarises conversations into minutes or notes with tasks
Language and dialect recognition in asylum procedures
Infers a person's origin or dialect from how they speak
Biometric verification and video identification
Uses a voice to verify a person's claimed identity
Data and information
Processing and operating environment
Recognition models usually come from providers; BenchNotes uses Azure Speech Services.
Storage and further use
Some deployers adapt recognition using their own recordings or specialist vocabulary. Corrected transcripts can feed back as new training examples.
Results and data sharing
Cloud services send audio recordings outside the organisation's environment.
Using a provider does not automatically determine where processing takes place. Storage location, subcontractors, deletion periods and use for training must be checked for the specific deployment.
Assess risks and determine measures
From understanding the system to analysing risks
- Define the system boundary
Record data, participants, interfaces and permissions. Resolve uncertain characteristics in the model.
- Investigate threats
Relate review candidates to actual data flows: what could happen, and who would be affected?
- Assess risk
Assess impacts and likelihood in the deployment context. The number of review candidates is not a risk score.
- Provide evidence of measures
Select appropriate controls, assign responsibilities and define effectiveness checks. Document residual risks and approvals.
Key threats in the application context
- 3 particularly important
- 5 typical
- 7 depending on the system
- 8 additional threats for every AI system
A total of 23 threats from the threat catalogue may be relevant to this use case.
Priority review points
Factual Inconsistencies (Hallucinations)
During a pause or unclear speech, the model inserts a sentence never spoken. In a finding or decision, it is later treated as the person's statement.
The transcript reads smoothly and looks complete. Approval without checking the recording misses misheard figures, names and negatives.
Sensitive Information Disclosure
Dictation and conversations contain health, case or account information. Provider retention or overly broad sharing exposes recordings and transcripts to unauthorised people.
5 further review candidates in the typical design
Other relevant threats
Targeted Poisoning / Label Poisoning
Falsified training labels can deliberately shift model outputs.
Listed because of this characteristic
- Learns from examples with correct answers
Training data or an adopted base model can introduce hidden triggers for misbehaviour.
Listed because of this characteristic
- Uses another provider's model
Complex or high-volume inputs can exhaust model computing resources.
Listed because of this characteristic
- Generates its own text
Targeted queries can recover some personal or protected training data from the model.
Listed because of these characteristics
- Processes personal data
- Processes confidential data
Exfiltration from ML Application
Protected data can leak through tools, network connections, storage or insiders.
Listed because of this characteristic
- Processes confidential data
Up to 7 further threats may apply, depending on your system design. Nine threats need to be reviewed for every AI system. One of these is particularly important here and is therefore listed above.
System features and open questions
Each Yes adds further threats to your checklist, up to 7 in total. If you cannot yet answer a question, include its threats in your review.
- Added by this Yes
- Also requires a Yes to another question
How your system is built
- Learns from your own data1 threat, plus 1 only in combination with other characteristics
- Uses operational feedback1 threat
How you operate the system
- Accepts queries directly through an interface4 threats
- Available to people outside the organisation4 threats
- Accessible over a network1 threat, plus 1 only in combination with other characteristics
- Incurs costs with each query1 threat
- Shares computing resources with other customers or applications1 threat
- Used alongside unapproved AI services1 threat
How your system is built
Does your organisation train or fine-tune the model itself using its own data?
Some deployers adapt recognition using their own recordings or specialist vocabulary.
If yes, add
Does operational feedback or output later feed into analysis or training?
Corrected transcripts can feed back as new training examples.
If yes, add
How you operate the system
These questions arise for almost every AI system.
Can people or programs query the model directly, for example through an interface?
If yes, add
Is the system accessible over a network or dependent on network services?
If yes, add
Do you pay per query or unit of computing time, so costs increase with use?
If yes, add
Does the system share servers or computing resources with other customers or applications?
If yes, add
Can people outside your organisation use the system, including without login?
If yes, add
Do staff also use AI services for this task that your organisation has not approved?
Staff sometimes upload recordings to freely available services when no internal tool exists.
If yes, add
Controls and implementation
Measures for the priority review areas are expanded. Each linked measure explains how to implement it.
Review priorityFactual Inconsistencies (Hallucinations)3 measures
Review priorityOverreliance3 measures
Review prioritySensitive Information Disclosure3 measures
Targeted Poisoning / Label Poisoning3 measures
Backdoor ML Model2 measures
Model Denial of Service3 measures
Privacy Attacks3 measures
Exfiltration from ML Application3 measures
Controls for the remaining threats are available on their pages in the threat catalogue.
AI Act provisions
Applies to every AI system
Art. 4: AI literacy
Applies if staff or appointed representatives operate or use AI systems.
Applies since 2 February 2025.
Obligations and details for Art. 4: AI literacy
- Who bears the obligations
- Provider, Deployer, Product manufacturer, Provider of a general-purpose AI model
Detailed obligations
Generally does not apply
Art. 50 para. 2: Marking artificially generated content
The transcript reproduces speech and generally creates no new content.
Obligations and details for Art. 50 para. 2: Marking artificially generated content
- The general condition is
- Applies if the system generates synthetic text, images, audio or video.
- Exception
- Exceptions include ordinary editing assistance that does not substantially alter inputs or their meaning.
- Points to consider
- Systems marketed before 2 August 2026 must be updated by 2 December 2026.
- Who bears the obligations
- Provider, Product manufacturer, Provider of a general-purpose AI model
Detailed obligations
Classify your system with the quick checker All obligations in the obligations compass
Cross-cutting threats
Review these threats for every AI system, whatever its purpose.
Atypical system features
View 18 further questions
If any of these features does apply to your system, add the listed threats.
Does the system generate or edit images, sounds, voices or videos?
If yes, add
Does the system create new content from open-ended requests that people read, see or hear directly? Translation, reading aloud and transcription do not count.
The system transcribes what was said and creates no new content.
If yes, add
Can users give the system natural language tasks that it then performs?
The system accepts no instructions; it only transcribes.
If yes, add
Does the system search its own collection of documents or records for relevant content?
If yes, add
Does the system formulate answers from retrieved documents?
If yes, add
Does the system use a knowledge graph storing concepts and their relationships?
If yes, add
Does the system remember content from earlier conversations or tasks and reuse it later?
If yes, add
Does the system call other programs, interfaces or search services itself, even if only to read?
If yes, add
Can the system change things itself, such as storing data, sending messages, initiating transactions or granting permissions?
If yes, add
Does the system determine its next steps itself based on earlier results?
If yes, add
Does the system delegate tasks to other AI agents or incorporate their results?
If yes, add
Is the system connected to tools or data sources through the Model Context Protocol (MCP)?
If yes, add
Do documents from several customers, departments or permission groups share one search store?
If yes, add
Does the system directly move or control something physical, such as motors, valves, vehicles or temperatures?
If yes, add
Does the system process several input types together, such as text with images or speech?
If yes, add
Does the system generate code, scripts or database queries for further use?
If yes, add
Does the system execute programs, scripts or commands itself?
If yes, add
Is the system intended to maximise a specified metric, such as revenue, utilisation or clicks?
If yes, add
Sources
- Statistics
Federal Statistical Office (Destatis), AI-using enterprises by technology type, 2025
Among AI-using companies, 42 per cent use speech recognition.
- Register or case collection
UK Algorithmic Transparency Record, HM Courts and Tribunals Service BenchNotes
BenchNotes automatically transcribes judges' dictation.
- Report of a deployment
Federal Employment Agency, List of AI applications, 26 January 2026 (published via Table.Media)
The agency converts speech into text for documentation.
- Authority, court or parliament
BaFin, Risks in Focus 2026: Digitalisation
BaFin reports smaller institutions automatically transcribing customer conversations.
- Specialist study
Koenecke et al., Careless Whisper: Speech-to-Text Hallucination Harms (arXiv 2402.08021)
In roughly one per cent of examined transcripts, Whisper invented complete sentences absent from the recording.
Related use cases
Also in the field of Office work and knowledge management
- Generative AI assistant for staffStaff draft, revise, translate and shorten texts using a language model their organisation provides in a secured environment.
- AI assistant in office applicationsStaff invoke the assistant directly in email, word processing, spreadsheets and chat; it accesses mailboxes and file storage using their permissions.
- AI knowledge assistant (enterprise search)The assistant answers questions from a defined document collection and shows which documents support the answer.
- AI meeting minutes and conversation documentationFrom recorded meetings and consultation conversations, the system drafts minutes or a case note with an action list.
- Machine translationPublic authorities and businesses automatically translate texts, websites and documents, often using their own glossary.
- AI agentAn AI agent breaks a task into steps and completes it largely independently using tools, interfaces and its own permissions in business applications.
Often used together with
Updated: 30 September 2026. Review checklist rules version 0.3.2, threat catalogue version 2026.10.24.2. How these pages are structured