Experiment with dAIsy

1. Purpose

This policy sets out how dAIsy may be used safely, responsibly, and cost-effectively during the University’s proof-of-concept (PoC) phase.

It complements, and does not replace, existing Imperial policies on academic integrity, assessment, research ethics, information security, GenAI, and data protection. Where there is any conflict, University policy prevails.

The policy aligns with Imperial’s guidance on appropriate use of generative AI and associated principles for teaching, learning and assessment. 

2. Definitions 

Generative AI (GenAI): Systems that produce text, code, images, etc., in response to prompts. 
Prompt / Output: The input you type (and files you upload) / the system’s response. 
Personal data / Sensitive data: Data that identifies a living person or special-category data (e.g. health, ethnicity). 
Academic misconduct: Use of GenAI that breaches University assessment rules (e.g. presenting AI-generated work as your own). 

3. Roles and access

  • Basic User Access - Exists by default for everyone with access to dAIsy. This access level allows for the creation of 'Personal agents'.
    • Personal Agents - Created by user for personal use
  • Developer  User Access – Developer access allows users access to special use case agents for a specific audience.(Specialised Agents are developed either by ICT or by a user who has been granted Developer Access). ICT will review and approve requests for developer access. If you require developer access, please complete the AI User Cases form.
  • Admin User Access - Strictly for ICT team. Access level allows for configuring roles, monitoring cost, and managing incidents.  

Note – Full role matrix is maintained separately and may change during the Proof of Concept (PoC) period. 

4. Acceptable use

  • Draft, rewrite, and improve non-confidential content; explore ideas; summarise provided materials.
  • Support learning and teaching with proper acknowledgement where required by your course/department; follow guidance on referencing AI use.
  • Prototype task-specific chatbots (Creators) within assigned caps and policies.
  • Prepare non-sensitive training, FAQs, and communications materials. 

5. Responsible uses 

Staff and students can use dAIsy as you would use other Imperial approved applications and services regarding the use of data. For example, you can input data into dAIsy as you would when using applications in your Imperial Microsoft 365 account. Review our university policies and guidance for further information on the acceptable use of data and the Code of Practice 6 - Conditions of Use of IT Resources.

i) Assessment-related use- Individual departments may allow or prohibit GenAI for specific assessments; local (team/department/faculty) instructions take precedence. Always check the brief and acknowledge permitted AI use as directed. 

ii) Academic integrity and referencing - Where GenAI use is permitted for assessed work, acknowledge it (e.g., tool, version, prompt approach). Improper use can be investigated under University Academic Misconduct procedures [PDF]. 

iii) Research use - Research that involves people, personal data, or sensitive topics may require ethics approval, a Data Protection Impact Assessment (DPIA), and data-governance controls before using any AI tool.

6. Data handling, security and privacy

  • dAIsy uses University SSO and auditing. Logs (prompts/metadata) are retained for operational monitoring. Model providers are configured to not train on your data.
  • Outputs may be inaccurate or biased, you must verify facts and cite sources where appropriate.

7. Cost and usage controls 

  • Token, cost and model usage is monitored. Caps are enforced per role to protect budget and ensure fairness.
  • Administrators may 'throttle' or pause access if costs or misuse risk exceed thresholds. 

8. Quality, bias and human oversight

Always apply critical judgement. GenAI can ‘hallucinate’, omit context, or reflect training-data biases. Users remain accountable for the accuracy, legality, and appropriateness of any content they submit or share.

9 Training and support

10. Logging and incident reporting

Usage is logged for security, cost, and service improvement. Report security incidents immediately via the ICT Service Desk (For tickets use category: Generative AI).

11. Copyright and IP

Do not upload third-party content you are not permitted to share. Ensure that any reuse of AI outputs complies with licensing and academic citation norms. 

12. Communications and transparency

When communicating externally, do not present dAIsy outputs as Imperial’s position without approval. For assessed work where AI use is permitted, be transparent about how you used AI. 

13. Compliance and sanctions

Breaches may lead to action under Academic Misconduct (students) and HR/disciplinary processes (staff), as well as Information Security and Data Protection policies. Sanctions may include removal of access, grade penalties, or formal disciplinary measures. 

14. Change control and versioning 

This policy will be updated during the PoC as capabilities and University guidance evolve.