Sinéad Gibney (Dublin Rathdown, Social Democrats)
Link to this: Individually | In context

1254. To ask the Tánaiste and Minister for Justice and Equality the risk assessment or bias testing was carried out on public facing AI chatbots used by his Department (details supplied) in line with the Guidelines for the Responsible Use of AI in the Public Service and Public Sector Human Rights and Equality Duty; and if he will make a statement on the matter. [57800/25]

Sinéad Gibney (Dublin Rathdown, Social Democrats)
Link to this: Individually | In context

1255. To ask the Tánaiste and Minister for Justice and Equality the procedures in place to ensure that chatbot used by his Department in dealing with the public are not biased and do not provide incorrect or misleading information; and if he will make a statement on the matter. [57801/25]

Sinéad Gibney (Dublin Rathdown, Social Democrats)
Link to this: Individually | In context

1256. To ask the Tánaiste and Minister for Justice and Equality the reason public facing chatbots being used by his Department were have not been subject to public tender processes; and if he will make a statement on the matter. [57802/25]

Sinéad Gibney (Dublin Rathdown, Social Democrats)
Link to this: Individually | In context

1257. To ask the Tánaiste and Minister for Justice and Equality to provide the names of all vendors assisting in the use of AI chatbots by his Department; and the amount they have each received for their services, in tabular form. [57803/25]

Jim O'Callaghan (Dublin Bay South, Fianna Fail)
Link to this: Individually | In context

I propose to take Questions Nos. 1254 to 1257, inclusive, together.

My Department’s use of AI-enabled chatbots is part of a broader transformation of customer services in the immigration area. The approach taken is low-risk, transparent, and in line with national AI and procurement guidelines. AI tools are used to enhance efficiency and accessibility but do not make decisions affecting individual rights. Continuous monitoring, human oversight, and clear disclaimers ensure responsible and ethical use.

The Tara (ISD) and Erin (IPO) chatbots did not use Generative AI. These chatbots were programmed with pre-scripted information, following a governance process in the organisation, to show applicants how to engage with our services and answer frequently asked questions. Tara was replaced with more updated technology as part of the Digital Contact Centre project introduced in October 2024. Erin was retired in Q2 2025. These chatbots were supported and maintained by a company called Ludex, procured through a central Government Procurement Contract .

In 2023, my Department issued a public Request for Tender for a cloud based customer service solution comprising of a digital contact centre, real-time online conversation assistance, an appointment booking system, and self-service system for customers to interact with our provided services. The deployment of the Digital Contact Centre incorporated a chatbot using Microsoft Copilot to provide assistance to both applicants and staff. It also incorporates a service, ID-Pal, for identity verification. These tools are part of an overall larger IT system. My Department has fully complied with public procurement guidelines in the acquisition of these services.

The chatbot implemented within the Digital Contact Centre operates within an authenticated and secure customer portal and has been trained only on information published on the Immigration website. This ensures that responses are accessible and based solely on verified, up-to-date, and publicly available content. Users are clearly informed that they are interacting with an AI system and are advised to verify the information provided. This transparency aligns with obligations within the EU Artificial Intelligence Act, under which chatbots are considered “limited risk” systems. The chatbot is designed to provide general guidance and does not provide legal advice. Customers can contact a human service representative at any stage.

Ongoing monitoring, staff training, and performance evaluation (including customer satisfaction scoring) are in place to ensure the chatbot continues to provide an accurate and equitable service.

In line with public procurement and AI governance guidelines, my Department undertook a comprehensive risk assessment process as part of the deployment of its Digital Contact Centre (DCC).

As part of the development phase, the chatbot output was tested extensively by my Department's customer service team members. This testing is maintained after 'go live' to ensure our service meets the needs of our users. The chatbot was trained solely on publicly available information from the Immigration website and does not make any automated decisions or determinations that affect individual rights or entitlements. This approach aligns with the Guidelines for the Responsible Use of AI in the Public Service and the Public Sector Human Rights and Equality Duty by ensuring the system is transparent, proportionate, and fair.

My Department also carried out a Data Protection Impact Assessment on the full IT solution prior to deployment, in consultation with the Data Protection Officer, to ensure that data is handled securely and privacy rights are protected. Bias mitigation was further supported through diverse testing teams and user design principles during the development phase.

The following vendors have been involved in assisting with the use of chatbots in the Department.

EY were contracted to work on the overall Digital Contact Centre, and the cost of the chatbot cannot be extracted from the overall cost of the Digital Contact Centre project.

My Department plans to continue to assess the opportunities presented by all digital tools, including AI, to improve the delivery, security, efficiency, and accessibility of services to our customers. Approval for adoption of any new tools or solutions is subject to appropriate governance processes, including risk assessment. An AI policy and governance approach, in line with national guidelines, is being drafted currently.