Acceptable Use of Artificial Intelligence Policy

This policy outlines guidelines for the responsible and ethical use of artificial intelligence (AI) technologies within [name of organisation]. It is designed to ensure that AI is used in a way that aligns with our values, protects sensitive information, and promotes innovation without compromising ethical standards. The policy aligns with [name of organisation’s] values in this way: [include organisational values relevant to use of AI]

The purpose of [the organisation]'s Acceptable Use of AI Policy is to guide the ethical and responsible use of AI technology by staff, volunteers and contractors at [name of organisation]. It aims to protect data privacy, ensure compliance with legal and regulatory requirements, and foster accountability. By providing clear guidelines on data handling, consent and security, the policy seeks to maintain the integrity of [name of organisation’s] operations while using the opportunities that AI offers.

This policy applies to all employees, contractors, and volunteers who use AI technology within the organisation. It covers the use of AI for content creation, data handling and any other relevant applications. Refer to the Critical Incident Response Procedure [if you have one] for guidance on how to respond to a data breach involving the use of AI.

Artificial intelligence (AI): The simulation of human intelligence in machines that are programmed to think and learn like humans.

Large language model: AI systems such as ChatGPT that can produce human-like text based on the input they receive.

Confidential data: Any organisational or client information that is sensitive and not intended for public disclosure.

Machine learning: A subset of AI that involves the development of algorithms that allow computers to learn from and make decisions based on data.

Algorithmic bias: Systematic and repeatable errors in AI systems that result in the unfair treatment of certain groups, often due to biased training data or flawed algorithm design.

AI systems: As a non-exhaustive list of examples, [Name of organisation] uses the following AI systems: [e.g. chatbots, predictive analytics tools and large language models].

Legal and regulatory compliance: Ensure that AI use complies with relevant laws, regulations and industry standards.

Ethical standards: Practise and promote the ethical use of AI, emphasising respect, transparency, and accountability.

Data protection: [Name of organisation] is committed to protecting individuals' privacy and maintaining trust by ensuring that personal and confidential data of beneficiaries, staff, volunteers and clients is handled responsibly and ethically when AI applications are used.

Confidentiality guidelines: Adhere to confidentiality protocols when handling personal and confidential data in AI applications, ensuring that data is stored securely when inputting it into AI systems, accessed only by authorised personnel, and encrypted where possible and appropriate.

Data classification: Classify data based on its sensitivity and apply appropriate protection levels. Sensitive data such as personally identifiable information and confidential business information requires the highest level of protection, meaning only staff with relevant areas of work should have access to the data.

Managerial approval: Obtain consent from a manager before using organisational or client data in AI applications.

Informed consent from external users: For AI systems that collect data from external users (e.g., customers, clients), obtain informed consent. This involves clearly informing users about what data will be collected, how it will be used, and their rights regarding their data. [Name of organisation’s] Privacy Policy can be found here: #####

Principle of least data: Adhere to the principle of data minimisation, which entails collecting only the data that is strictly necessary for the specific AI application.

Purpose limitation: Ensure that data is used only for the purposes explicitly stated at the time of collection. Any secondary use of data requires additional consent and must be clearly justified.

Data anonymisation: Wherever possible, use anonymised and aggregated data to minimise privacy risks. Anonymisation involves removing personally identifiable information, while aggregation involves combining data sets to prevent the identification of individuals.

Algorithmic bias identification: Analyse AI systems for potential biases in training data, model design, and decision-making processes.

Mitigate algorithmic bias: Implement strategies to correct identified biases, such as adjusting prompts or raising awareness in the team.

Handling confidential data: Do not input confidential organisational or client data into any AI program without consent from a manager. This is to guard against possible breaches of confidentiality and to maintain the integrity and security of sensitive information.

Data use settings: Ensure that all settings on the AI language models used are configured to "do not use data to inform machine learning." This setting is crucial to protect the privacy and confidentiality of the data being processed and to prevent it from being used to train AI models further.

Human oversight: All outputs generated by AI systems must be reviewed and verified by a human before being used in any official capacity.

Correction of errors: If AI-generated content contains inaccuracies or errors, it is the responsibility of the user to correct these mistakes before dissemination.

Original content creation: Avoid copying and pasting text directly from AI-generated content. AI may be used to generate, inspire and form ideas, and improve writing style, but the final output must be a product of your own creativity and expertise.

Attribution and transparency: When AI large language models significantly influence the creation of content through the direct provision of text or imagery, you must clearly acknowledge the use of AI in your work.

Ethical considerations: Use AI only in a manner that aligns with the organisation’s values.