OTTAWA — When the Ontario Provincial Police began using Palantir Technologies’s controversial Gotham database product in 2015, it was an “operational issue” and didn’t cross the desk of the minister responsible for them, said a spokesperson for that minister, Ottawa politician Yasir Naqvi.

The Logic revealed in the fall that the OPP has had a contract with U.S.-based Palantir for its “analytical platform,” after pursuing details through access-to-information law for nearly a year. After publication, the Ontario government supplied one additional detail: the Palantir contract began on June 1, 2015.

Naqvi is now a federal Liberal MP and parliamentary secretary to Bill Blair, the emergency preparedness minister. He’s also contemplating a run for the provincial party leadership.

At the time, though, he was Ontario’s community safety minister under Liberal premier Kathleen Wynne, and overseeing the province’s creation of new rules on carding—the police practice of randomly stopping people to ask for their identity papers.

Typically, that data goes into systems like Gotham, which is billed as a tool to fight wars, solve crimes and even predict offences that haven’t been committed. The point is to use software to make connections that humans wouldn’t.

Critics of that technique of mass data collection said—and still say—Naqvi’s reform didn’t go far enough to keep police from gathering information improperly.

“The law is so full of holes, or is being interpreted to be so full of holes, that [police] are basically able to do what they were doing before,” said Abby Deshman, the director of the Canadian Civil Liberties Association’s criminal-justice program, in an interview.

In practice, “random” checks have disproportionately targeted racialized minorities and Indigenous people, and a provincial review concluded that they’re of negligible crime-fighting value. Advocates for racialized people, including Ontario’s chief human-rights commissioner Renu Mandhane (who is now a judge), criticized Naqvi’s regulation at the time for leaving too much police intelligence-gathering unrestricted. 

Naqvi lost his provincial Ottawa Centre seat to a New Democrat when the Progressive Conservatives under Doug Ford blew the Liberals out in 2018, but ran successfully for federal office in 2021. 

The Palantir contract “would have been an operational issue that would not have involved the minister at the time,” Naqvi’s current legislative assistant, Jared Maltais, told The Logic by email.

That police store information gathered from carding stops and other interactions in databases for later analysis was part of the consultations Naqvi led at the time. The regulation that resulted explicitly allows it.

Palantir is a multibillion-dollar artificial intelligence company based in Colorado, co-founded by the hard-right investor Peter Thiel. It has massive government business—including some contracts that other Big Tech players avoid and a critical role supporting Ukraine in its fight against Russia. 

Naqvi also did not know about the use, previously reported by The Logic, of a legal provision that shielded the Palantir agreement from routine budget scrutiny as an expense related to solving crimes, Maltais said. Using it required sign-off from either Ontario’s attorney general or her deputy. (Naqvi moved on to become attorney general himself in 2016.)

“Ending the discriminatory and arbitrary ‘street checks’ was one of the early reforms I brought forward to cabinet,” Naqvi said in a statement that Maltais relayed. “We also introduced many much-needed policing reforms that were swiftly reversed by the Ford Conservatives.”

Naqvi’s regulation on arbitrary street checks says police can’t be “blatantly discriminatory,” Deshman said, but doesn’t stop the collection of large quantities of data.

 

The OPP reported seven “interactions” covered by Naqvi’s regulation in 2020, the last year for which they’ve published an annual report. That’s down from 94 in 2017, the first year the new rules applied.

In 2018, though, the review of the new regulation pointed out the huge number of police dealings where the carding regulation doesn’t apply. Those include traffic stops and enforcement of provincial laws on liquor and trespassing; and “observation checks,” which are when police record seeing someone without stopping them to request their papers. None of those needs to be reported.

Justice Michael Tulloch, the same judge who had delivered the first report saying carding is bad policing (and the new chief justice of Ontario, as of last December), acknowledged that on paper, it appeared the number of these interactions had plunged, but he warned not to take that to the bank.

“A reason for the sharp decline in the numbers of what are commonly referred to as street checks post-regulation is that the numbers outlined pre-regulation, which often were in the thousands, included both regulated and non-regulated interactions grouped together under the street checks module,” he wrote in his report.

In other words, there could still be thousands of unreported interactions feeding databases in police forces across Ontario, fully in keeping with the rules.

Police weren’t adequately trained on the regulation or the reason for it, Tulloch found, and the rules didn’t meaningfully restrict the use of the information once it was collected.

The uncertainty about what happens to that data when it’s fed to analytics tools is troubling, said Brenda McPhail, the director of the CCLA’s work on privacy, technology and surveillance, who spoke to The Logic along with Deshman.

The province refused to say just how the OPP uses Gotham, citing a need to protect its investigative techniques. Nor would it disclose how much the contract is for, saying it intends to put a new contract up for bids soon and doesn’t want to reveal what it’s paying Palantir now. Palantir did not reply to questions about its business with Ontario.

McPhail said that lack of information erodes public faith in policing. “We have had a series of revelations about different policing bodies across Canada, using different surveillance tools, where the way we find out is from a media story, or an intrepid reporter gets a lead and follows it up, rather than from the force. So these products are being considered, procured and used mostly in secret,” she said.

Police services started using New York-based Clearview AI’s facial-recognition software without consulting privacy authorities, she pointed out. When the federal privacy commissioner at the time investigated, he found that Clearview built its analyses on illegally compiled images.

Stronger rules, based on a public debate, about how police acquire and use tools that rely on big data might have led the departments involved to ask about that before starting to use Clearview AI’s product. 

“Before we make a decision to actually use one of these products, that sort of public consultation needs to be baked into an effective and rights-respecting procurement process,” McPhail said.