The Ada Lovelace AI research institute has released guidance for public sector organisations on the use of foundation AI models. The policy document, titled “Foundation models in the public sector,” outlines the considerations for deploying these models in a variety of applications, such as translation, report generation, and query response. The Ada Lovelace Institute’s mission is to ensure that data and AI benefit society, and it has been vocal about the government’s approach to AI and ethics.
The release of the guidance comes after reports that the civil service is using AI chatbots to replace human workers. The Public and Commercial Services Union (PCS) has raised concerns about AI deployment in the public sector, urging appropriate regulation to avoid discrimination against staff. The union believes that the guidelines issued by the Ada Lovelace Institute are not enough and that legislation and collective agreements are necessary to protect workers.
According to the guidance, the use of foundation models in the public sector is currently inconsistent across government offices. While some civil servants and local authorities are informally using these models, their authorized use is limited to demos, prototypes, and proofs of concept. However, there is optimism in the public sector about the potential for these models to enhance public services and meet the growing needs of users within budgetary constraints.
The guidelines also highlight the risks associated with using AI foundation models, noting biases, privacy breaches, misinformation, security threats, overreliance, workforce harms, and unequal access as potential concerns. The Ada Lovelace Institute advises public sector organizations to improve their governance of these systems and consider risks when developing their own models or procuring external ones.
To ensure effective use of foundation models, the organization suggests regular reviews and updates to guidance, setting procurement requirements that uphold public standards, piloting new use cases before widescale implementation, holding data locally, and providing training to employees. However, the PCS union argues that guidance alone is insufficient and calls for legal regulation and collective agreements to protect workers’ rights and ensure that the benefits of AI are shared.
The release of the Ada Lovelace Institute’s briefing document and the concerns raised by the PCS union highlight the need for careful consideration and regulation of AI deployment in the public sector. It is crucial to address the risks associated with AI, protect workers’ rights, and ensure that the benefits of AI are harnessed for the benefit of society as a whole.