Canada’s border agency plans to use AI to screen everyone entering the country — and single out ‘higher risk’ people

Inevitable given the numbers involved and the need to triage applications:

Canada Border Services Agency is planning to use AI to check everyone visiting or returning to the country to predict whether they are at risk of breaking the law. 

Anyone identified as “higher risk” could be singled out for further inspection.

The traveller compliance indicator (TCI), which has been tested at six land ports of entry, was developed using five years of CBSA travellers’ data. It assigns a “compliance score” for every person entering Canada. It will be used to enforce the Customs Act and related regulations.

The AI-assisted tool is expected to launch as early as 2027 and is meant to help border services officers at all land, air and marine ports of entry decide whether to refer travellers and the goods they are carrying for secondary examination, according to an assessment report obtained under an access to information request. 

“We use the obtained data to build predictive models in order to predict the likelihood of a traveller to be compliant,” said the report which was submitted by the border agency to the Treasury Board.

“TCI will improve the client experience by reducing processing time at the borders. The system will allow officers to spend less time on compliant travellers and reduce the number of unnecessary selective referrals.”

However, experts are alarmed by the lack of public engagement and input into the tool’s development. They worry that the system may reinforce human biases against certain types of travellers such as immigrants and visitors from certain countries because the quality of the analytics is only as good as what is inputted.

“If you’ve historically been very critical over a certain group, then that will be in the data and we’ll transfer that into the tool,” said Vancouver-based immigration lawyer Will Tao, who obtained the report.

“You look for the problems and you find problems where you’re looking, right?”

The government report said the border agency serves more than 96 million travellers a year, and trying to keep up with expected growth would require the addition of hundreds of border officers. In addition, physical limitations make it impossible to add extra booths at some points of entry.

The AI tool, the report said, will help keep border processing times at current levels even with an expected increase in the number of travellers. 

“No decisions are automated,” the report said. “Rather the current primary processing is being supported with a flag indicating whether a traveller’s information matches a compliance pattern.”

However, if an officer follows a mistaken recommendation from the tool, it could have impacts that could “last longer,” the report added.

“Once a risk score or indicator is presented to an officer, it can heavily influence their judgment, which in practice means the system is shaping outcomes even if the final authority is technically still human,” said University of Toronto professor Ebrahim Bagheri, who focuses on AI and the study of data and society.

“A false positive is when the system flags someone as risky or non-compliant even though they are in fact compliant. In the border context, that could mean a traveller is singled out for extra questioning or secondary examination even though they’ve done nothing wrong.”

The system is designed to display information of interest to an officer, such as a traveller’s means of transport and who accompanied them. 

It also captures “live determinants” which can include information such as whether the person is travelling alone, the type of identification they presented and the license plate of the vehicle they used, as well as any data from the traveller’s previous trips in CBSA’s records….

Source: Canada’s border agency plans to use AI to screen everyone entering the country — and single out ‘higher risk’ people