Efforts to explore how artificial intelligence (AI) might support parliamentary work are being led by a vanguard of between 20 and 30 parliaments. “Early adopters” of AI in the world of parliaments include Brazil, Canada, Denmark, Estonia and Italy, as well as the European Parliament. But parliaments in all regions are playing close attention to developments in this area. The Southern Africa Regional Hub, for example, organized an online event on AI in parliaments in March 2024.

Within the Inter-Parliamentary Union’s (IPU) Centre for Innovation in Parliament (CIP), senior technical experts regularly come together to share their experience and develop good practices. Through this community, two AI-related knowledge products have been developed so far: an online event and an issue brief. The community is now working together to co-create Guidelines on AI governance in parliaments for publication in the second half of 2024.

Online event

The online event on AI in legislative processes, held in October 2023, drew out a number of lessons learned.

Mr. Christopher Henry (House of Commons of Canada) outlined ways in which AI can make existing processes more efficient. He explained that it was becoming increasingly difficult to find, train and retain stenographers, transcribers and other staff performing repetitive tasks, adding that pressure was increasing to deliver services faster and at higher quality. He argued that AI was therefore arriving at the right time, noting the importance of close cooperation between business departments and IT staff in order to find ways that AI could add value in parliamentary processes.

Mr. Carlo Marchetti (Senate of Italy) discussed the notion of parliaments partnering with universities in order to develop proofs of concept and prototypes. He emphasized that parliaments nevertheless needed to retain full responsibility for the systems they use. He also stressed that, following the proof-of-concept stage, the technology should be transferred from the university to the parliament. Mr. Marchetti also touched on the subject of generative AI (GenAI) systems, noting the need to carefully consider the implications associated with their use.

Issue brief

Following this exchange of experience, the CIP drew on the knowledge within the community to prepare an issue brief entitled Using generative AI in parliaments. Published in March 2024, it provides guidance at a time when parliaments are investigating use cases for ChatGPT and other GenAI systems, such as for summarizing committee meetings. However, these tools are “black boxes” and users have little or no control over the internal workings. Moreover, outputs from GenAI need to be carefully compared with what is currently produced by parliamentary staff in order to ensure the necessary level of quality. 

The key messages in the issue brief are summarized below:

• A step-by-step, risk-based approach will be appropriate for most parliaments. 
• Based on current parliamentary practice and research, parliaments are encouraged to begin experimenting with GenAI, for example, to save time with tasks such as producing summaries of texts and creating records of parliamentary debates. Such experiments should be carried out in a safe “lab environment” before being rolled out more widely.
• However, parliaments need to be extremely cautious about introducing GenAI in core legislative systems at this stage. High levels of confidence in the technology and strong governance processes are needed before such a step is considered.

The publication is available in English, French, Spanish and Arabic. It was presented to the Association of Secretaries General of Parliaments (ASGP) in March 2024 by the CIP and by Mr. Remco Nehmelman, the Secretary General of the Senate of the Netherlands.

Given that the technology is evolving rapidly, the guidance is expected to be updated regularly. Parliaments are invited to share feedback and updates with the CIP at [email protected].