Published: January 2025

This briefing note outlines some of the ethical considerations related to the use of generative AI that charities should contemplate when looking to use such tools.

1. Transparency, explainability and accountability

2. Data protection and copyright

3. Sustainability

4. Job security and inequalities

Transparency, explainability and accountability

The algorithms that underpin generative AI models are extremely complex. The way they arrive at a decision is not linear, making it difficult for even their developers to fully understand how they consider probabilities and make decisions to arrive at a final output based on the user input. This is known as the ‘black box’ problem.

The black box problem has implications for transparency and explainability of generative AI models, especially if they are used to assist decision making processes such as in screening job applicant CVs, where it is important to ensure fairness and lack of bias. It also makes it difficult for developers to understand and fix the sources of errors made by these models.  As with any other tool, it is generally accepted that it is individuals and organisations that are accountable for how generative AI is used.

One way to address these challenges is through a 'human-in-the-loop' approach, which emphasises collaboration between humans and AI systems. This interaction can take many forms, including explainable AI, where models clarify their decisions to human reviewers, and active learning, where systems remain under human oversight during the training process.

Users should consider these limitations and determine whether they are comfortable with using generative AI tools for specific tasks when the reasoning behind their decision making may not be fully transparent.

Data protection and copyright

Web scraping is the automated process of extracting data from public internet websites, including copyrighted materials and any personal information shared on social media sites or elsewhere online. Public chatbots may have been trained on vast amounts of scraped content, largely without explicit consent being sought from individuals and copyright holders whose information and materials were included in this public data.

In summer 2024, there were several ongoing international legal investigations looking to determine if this constitutes a breach of copyright or data protection regulations.

Sustainability

The world is facing increasing pressures related to climate change and environmental sustainability is a growing focus for many organisations. This creates a tension in this space, as the training and use of generative AI models has an environmental impact, requiring large amounts of electricity and water to power computers and cool data centres.

Users should consider the environmental implications and, when possible, choose developers demonstrating commitment to sustainable practices. Additionally, it is worth reflecting on whether generative AI tools are the most suitable option for a given task, balancing their benefits with their environmental footprint.

Job security and inequalities

Some are concerned that machines and AI will replace human jobs. However, most experts at the moment consider that whilst it will change the nature of some jobs, a more immediate threat is confident AI users potentially outperforming those who are less familiar with this technology. Larger organisations, with additional funding and resources, may also be better positioned to harness the opportunities presented by generative AI and digitally upskill their staff. As this field continues to advance, organisations of all sizes should consider how they best share knowledge and support their staff in learning how to use these tools effectively and responsibly.

Further information and resources

• Generative AI: Opportunities for charities
• Generative AI: Risks for charities
• Generative AI: Getting started
• Generative AI: Research application and assessment