How to create an AI policy

Artificial intelligence (AI) can’t be ignored. No matter how charities feel about it, the technology is here.  Part of being ready for AI is having a policy to determine what, when, and how the tech will be used.

 

Policy writing for AI is in its infancy, but is still worth doing to avoid pitfalls.

 

The Charity Commission sums it up broadly: “As use of AI develops and more applications become available, consider if having an internal AI policy would be beneficial so it is clear how and when it can be used in governance, by employees in their work, or in delivering services to your beneficiaries.”

 

Here we discuss the main things to consider when putting together your charity’s AI policy.

 

 

Define the objective

 

Organisations should consider what the aim is for using AI. Questions to answer include whether it’s to improve efficiency, deliver services, promote innovation, or to increase fundraising. Using AI for the sake of it may not be the best rationale.

 

 

Consider the ethics and risks

 

Ethics are a big issue for all AI users. This is especially true for charities given the high ethical standards that they are often held to.

 

When using AI to develop content, some of the key risks are:

• Biases. AI sweeps the internet for information, but this isn’t always the “neutral” viewpoint
• Plagiarism. When using AI, are creators even aware of what the sources are? Creators could be using the work of others without the proper citations
• Security. AI tools could be targeted by hackers. The policy needs to consider how to prevent that and what would happen if data or processes were leaked

 

Assess the regulatory and compliance environment

 

Be aware of copyright and GDPR when assessing the regulatory and compliance of using AI. The Charity Commission can issue warnings and fines related to breaches.

 

Equally, from a reputation standpoint, AI doesn’t always do the best job in verifying information. Be wary of fake news and false reporting. Fact-checking is a must.

 

 

Create example use-cases

 

For the policy to have teeth, include use-case examples. These are relevant for when AI is and is not appropriate. Then, iterate why the cases are acceptable or not so those outside of developing the policy understand how it works.

 

Catalyst sketches out how generative AI tools (like ChatGPT) are deployed when producing content. They note that some charities use the tech to come up with ideas, help create social media, and to convert complex literature into video scripts.

 

Organisations might also decide that AI is only to be used for internal analyses. Prostate Cancer UK and Moorfields Eye Hospital both use it for data analytics to assist in service delivery:

 

Prostate Cancer UK, in collaboration with UCL and Movember are working with AI to help prevent and treat cancer more effectively. The AI, called STAMPEDE, scours data samples to figure out which treatments are most successful based on survival and healthiness at the end of treatment. In practice, the AI helps doctors build customised treatments over standard approaches. 

 

Moorfields Eye Hospital are using a similar approach. Their AI helps to identify eye diseases and may be able to help predict other problems.

 

Last, when it comes to use-cases, our top tip is to include situations where not to use AI. Charities might want to highlight examples that jeopardise effectiveness, relationships and reputation.

 

 

Understand AI governance and accountability

 

Zoe Amar has published a twelve-section checklist with questions that are considered, if not answered. Among the most pressing topics are governance and accountability.

 

On governance, the checklist outlines how all other policies may be impacted by use of AI. These include cyber security, data protection and overall digital policies.

 

Accountability means being able to stand behind the choices made around AI. The checklist further explores how to have transparent AI use and procurement.

 

 

Learn, develop, and train

 

The AI policy should cover who is using the AI and what skills and qualifications are required for each use-case.

 

Training staff in the age of AI might feel challenging, but there are many resources that are delivered online. We suggest starting with the basics.

 

Coursera offers sessions across many levels and ambitions for learners. You could start with those that focus on machine learning and artificial neural networks to get the basics around how organisations use ChatGPT and chatbots.