Our AI Code of Practice
UPDATED SEP 2026
1. Purpose
The purpose of this Code of Practice is to guide the responsible and constructive use of artificial intelligence (AI) and particularly large learning models (LLMs) within the Blavatnik School of Government.
It applies to all members of the Blavatnik School community and is designed to provide clarity for students, faculty, visiting fellows, and staff. It should be read in conjunction with the University of Oxford's policies on data protection, research ethics, and academic integrity.
2. Scope
For the purposes of this Code, AI refers to systems that can perform tasks requiring reasoning, adaptation, or pattern recognition, such as predictive text, voice assistants, or image analysis tools. Generative AI is a category of tools capable of producing new content such as text, code, images, video, or audio based on large datasets and probabilistic prediction. Generative AI includes, but is not limited to, tools such as ChatGPT, Gemini, Claude, and Microsoft Copilot.1
1 Assistive AI should be distinguished from generative AI insofar as it denotes tools that support tasks without substituting for substantive intellectual effort, such as spelling correction, transcription, or formatting. The Code primarily concerns the use of generative AI in teaching, research, and administration.
3. Three Simple Rules
- You are responsible for all content you produce using AI and for verifying any AI-generated content you use in your work. You will always be accountable for the veracity and accuracy of work you present to others. Note that students may not use generative AI for summative assessments (further guidance is set out in the student handbook).
- You must always disclose when AI has been used to generate the words or thoughts you are sharing with others. We expect that many people will use AI to augment and assist research, analysis and communication. It is good practice to explain how it has been used for these purposes. For example, a short prefatory sentence might explain that "AI tools were used to conduct background research, as well as to check spelling and grammar, to suggest alternative word choices, and to format this draft" or "having written this in my native language, I used AI to translate my message".
- You must protect the confidentiality of data which should not be uploaded into an LLM or commercial system (eg information protected under GDPR, personnel information).
4. Further Guidance
AI can be used to include (eg where English is not a native language, to express ideas with greater clarity and confidence) and equally, it can embed the bias of data sets on which LLMs have been trained which needs recognising and countering.
AI should be used to assist and augment work but already there is evidence of people using AI to substitute for their own thinking. This weakens capabilities to reason, to challenge arguments, to read, to memorize and to deepen intellectual capacity and curiosity. As a community we need to work together to evolve ways of working, teaching, and doing research which strengthen brainpower and capability.
This is a fast-moving issue. We will need to keep working together to clarify how we expect our colleagues and students to use AI in their work.