AI System Intake Form
Questionnaire for suppliers when procuring AI
Use this form to request essential information from AI system suppliers. This helps assess risks, compliance and suitability.
- Send this form to the supplier prior to procurement
- Request written answers with substantiation
- Use the answers for risk assessment and decision-making
- Keep completed forms as part of documentation
A. General Information
Basic information about the system and supplier
Name of the AI system
Name of supplier / developer
Country of supplier
Version number of the system
Intended purpose / use case in government
Contact person supplier (name, email, phone)
B. Technical Specifications
Details about operation and architecture
Type of AI technology (rule-based / machine learning / deep learning / LLM / other)
Describe how the system generates decisions/output
What data was the model trained on? Describe the training data.
What are the measured performance indicators? (accuracy, precision, recall, etc.)
Is the model further trained after implementation (continuous learning)?
Where does the system run? (on-premise / cloud / hybrid)
If cloud: which cloud provider and in which region are the servers located?
C. Data & Privacy
D. Bias & Fairness
Download the complete form
Receive the complete AI System Intake Form with all questions directly in your inbox.
You will also receive our monthly updates on AI governance for governments.
Full sections after registration:
C. Data & Privacy
Data processing and privacy aspects
What data does the system receive as input?
Are personal data processed? If so, which categories?
Are special categories of personal data processed? (health, ethnicity, religion, etc.)
How long is data retained?
Is data shared with third parties? If so, with whom?
Is a DPIA (Data Protection Impact Assessment) available?
How is data security guaranteed? (encryption, access control)
D. Bias & Fairness
Measures against discrimination and unfair outcomes
Have bias tests been conducted on the model? If so, describe the results.
Which protected characteristics were tested? (age, gender, ethnicity, etc.)
What measures have been taken to mitigate bias?
Is the system monitored for discriminatory outcomes after implementation?
E. Explainability & Transparency
How understandable is the system?
Can the system explain why a particular output was generated?
How is explanation provided? (feature importance, decision path, etc.)
Is the explanation understandable for non-technical end users?
Is documentation available about the operation of the model?
F. Human Control
Possibilities for human oversight and intervention
Can users override the system output?
Is there a "kill switch" to immediately disable the system?
How are deviating decisions by human operators logged?
Which roles/functions have access to the system?
G. Maintenance & Support
Support, updates and service levels
What is the availability guarantee (SLA)?
How often are updates/patches rolled out?
How is the client informed about changes?
Is 24/7 support available? If not, what are the support hours?
What is the process for failures or incidents?
H. Compliance & Certification
Compliance with regulations and standards
Has the system been tested against the EU AI Act? What is the risk classification?
What certifications does the supplier have? (ISO 27001, SOC 2, etc.)
Is a declaration of conformity available?
How is future regulation handled?
I. Documentation
Available documentation and resources
Is technical documentation (model card / system card) available?
Is a data sheet about the training data available?
Is audit documentation available?
What additional documentation can be provided?
Disclaimer: This form is intended as a tool for AI procurement. Always consult your procurement specialists, lawyers and IT experts for complete assessment.
Based on best practices for responsible AI procurement
Version January 2025