You should consider privacy risks when engaging in any activity on digital platforms, and especially those that have been newly developed such as generative AI (Artificial Intelligence). 

It is important to remember that all generative AI tools have ‘learned’ at some point in the past from the data, interactions, and behaviours of internet users. Some tools have a ‘cut-off’ point of only consulting data from several years ago, while others continue to learn from recent sources. 

Copilot’s privacy policies are broadly in line with those that apply to other Microsoft products. Imperial is currently comfortable with Copilot’s approach to user privacy and the extent to which our staff and students can browse the tool safely and with relative anonymity. It is for this reason that Imperial has activated our institutional subscription and is offering our community various guidance on how to use this tool. 

We recommend you take a cautious approach when deciding whether to upload any sensitive data into any generative AI platform, including Copilot. Terms and conditions, platform ownership, server locations, and government legal frameworks can all swiftly change. This is a particular risk when an innovative technology is being developed and rolled out across different countries and sectors of society and the economy. 

If pre-existing policies might apply to the data you are seeking to analyse (e.g., NHS patient data, data relating to research participants, or data about Imperial students), we recommend you consult these policies before taking any further action. The potential advantages you may garner from using powerful generative AI tools do not outweigh data subjects’ reasonable expectations that their right to privacy and data protection be respected. 

It depends on the context. Ethics is a key part of the ongoing discussion as to how AI might affect our daily lives, and Imperial is considering these big questions too. We have no intention of banning the use of these tools, but we do ask our community to carefully consider when to use them. 

If you are an Imperial student and you are submitting assessed work for your degree programme, many departments have taken the view that any use of generative AI must be at least cited and, in some cases, explicitly permitted beforehand. Failure to do so will diminish your own understanding of your learning outcomes and is likely to constitute Academic Misconduct. You can find out more about Imperial’s approach by visiting the AI & Study Guidance Hub and by discussing your use of AI with the teaching team within your programme or department. 

If you are an Imperial staff member, or even a contractor working with the university to provide services, it is worthwhile to consider whether an Imperial colleague would expect you to have used generative AI when completing a task. Might they reasonably be under the impression that an output (e.g., a report, a design, or a video) has been produced by you and solely you? In this situation it would be improper not to cite the generative AI platform you have used.  

Policies are being developed across the university to give the clearest possible guidance for a variety of contexts in which generative AI might be deployed. These will take time to be approved and publicised. In the meantime, all members of the community should consult the Imperial Values and Behaviours and apply them to all interactions they may have across the university.