Department of Justice criticised for using AI to manage immigration status of applicants

Irish Council for Civil Liberties said shortcomings in system 'present a real risk of harm to some of the most vulnerable people in our society'

System includes an 'AI-enabled smart assistant and chatbot powered by Microsoft Copilot, which is being used to help applicants manage their application status and submit queries'.

The Irish Council for Civil Liberties (ICCL) has voiced concerns about possible "shortcomings" at the Department of Justice using artificial intelligence (AI) tools for people to manage their immigration status.

The tools have been rolled out within the Department of Justice’s Immigration Service Delivery section, with it operating a ‘Digital Contact Centre’

This includes an “AI-enabled smart assistant and chatbot powered by Microsoft Copilot, which is being used to help applicants manage their application status and submit queries".

The department said the AI tool is trained exclusively on information published on the Immigration Service Delivery website and it does not process or make decisions based on an individual’s personal data.

“More complex issues can be referred to staff at any stage,” justice minister Jim O'Callaghan said.

“In addition, the department uses ID Pal, which employs machine learning-based identity verification to authenticate applicants’ identity documents when establishing a customer account within the same service portal.”

However, the ICCL said it had learned via Freedom of Information requests no tender process or risk assessment had been carried out.

“These shortcomings present a real risk of harm to some of the most vulnerable people in our society,” an ICCL spokesperson said.

Chatbots are only one example of algorithmic systems. We do not know how many other algorithmic systems and which kinds are being used in Irish public bodies.

Mr O’Callaghan said his department complied with existing guidelines on AI use within the public service outlined by the Department of Public Expenditure, as well as guidance from the National Cyber Security Centre.

“My department is fully committed to transparency, including informing users whenever they are interacting with an AI system and has put in place appropriate governance, risk assessment and oversight structures to support the responsible use of these technologies,” he said.

The information was provided to Sinn Féin’s John Brady via parliamentary question.

Mr O’Callaghan said the Department of Justice did not currently use generative AI tools for responding to parliamentary questions, while adding it would “continue to assess emerging digital tools”.

He added the primary use of AI within the Department of Justice was for dealing with immigration issues.

The ICCL said the State should introduce a publicly accessible register to detail the use of algorithmic systems by public bodies.

Meanwhile, the presidents of the Court of Appeal and the High Court have issued directions on how generative AI should be used by legal practitioners in court documents, which come into force from Tuesday.

In those directions, they warn generative AI could produce responses and material that “appear plausible but are false or misleading”, could generate “fabricated” material or fictitious cases, or “inaccurately summarise genuine sources”.

It also warned it could produce “omissions or biases” based on how an AI tool was trained or the way it processes data.

The presidents warned court documents where AI is used must be verified by a human before they are used in proceedings, and legal practitioners who do not comply with the directions — for example, by citing non-existent sources — may face sanctions.