Members of the Council:
I write to ask you to reconsider your approval, at your last meeting, of Automated License Plate Readers (ALPR) at the approaches to the Mount Hope Bridge. While the goal of suicide prevention is laudable, recording and searching 450,000 license plates each month is a serious intrusion into the privacy of everyday Rhode Islanders and is out of proportion to the number of lives such an action would purportedly save.

The solution to preventing jumpers is nets, not surveillance. The inclusion in the recently approved state budget of funding to begin the process of adding suicide barriers to the state's bridges makes adding ALPRs particularly ill-timed and redundant. Using this as a justification no longer passes the test of "least restrictive means" test for objectives at odds with Constitutional rights to privacy. For this reason alone, I urge the Council to abandon this project.

Nor does it seem that the vendor, Flock Safety, considers suicide prevention to be part of their core mission. Their web site leads with "Reduce crime in your community by up to 70%" and "Protect against property crime, violent crime, stolen vehicles, and more." The one mention of suicide prevention is an unattributed anecdote about a single case in Tega Cay, SC. No news articles could be found in a web search to substantiate this solitary claim. If this really was a featured benefit of the system, one would expect there to be more robust support; a general web search turned up no news stories in which ALPR were mentioned in preventing suicides. The use of this technology appears to be a solution in search of a problem at best, and at worst, a pretext by which to insert this vendor's cameras into our community.

Flock's web site claims that their cameras "capture objective evidence, make it actionable with machine learning." As someone familiar with ML, I can categorically state that all such systems are inherently imperfect -- that's the *nature* of machine learning. ML systems use weighted probabilities to tune their recognition, and they are susceptible to bias introduced by the sample sets used to train them. The point is that such systems can not, by their essential nature, be trusted not to produce false positives. The consequences of a police intervention that goes wrong based on an erroneous hit are clearly a foreseeable risk.

Turning to the Portsmouth PD policy included in the backup for the meeting, section III(F) articulates that the ALPR hits may include (but are not limited to) "stolen car, wanted person, missing person, domestic violation protective order or terrorist-related activity." If the stated objective of this deployment is suicide prevention, where is that in the list?

In section V(B) the policy states that ALPR will not be used to record identifying features of a vehicle like color, bumper stickers, or unique details "unless for a purpose authorized under subsection (a)" -- but section A casts a very wide net indeed, including vehicles associated with wanted, missing, or endangered persons, stolen vehicles, vehicles that return an NCIC match, and vehicles associated with criminal investigations. So any vehicle that falls into one of those categories will have more than its plate recorded.

And the Flock FAQ, included in backup, is emphatic on what will be captured (page 2, paragraph 3): "The Falcon captures the make, vehicle type, color, license plate (full, partial, missing), state of the license plate, and the unique features of the vehicle, including damage and after-market alterations." Captured and stored for a month and searchable. This should make anyone driving over the Mount Hope Bridge uncomfortable.

Policy section V(B) makes a point of saying that the number of times a vehicle travels past the camera will not be tracked -- but, again, the exception in V(C) says that "suicide prevention" would make tracking allowable. Is that only in the case of an existing BOLO, or is this a prospective machine-learning sifting through the data to find multiple crossings within a time span? Without clarification, this is unacceptably muddy. Students attending Roger Williams University who live in Portsmouth will undoubtedly be erroneously flagged up by the system, which will both waste PD time and send a very unwelcome message.

In section V(H), the policy states that Flock will "ensure proper maintenance and security of data stored in their data towers." As a career IT professional, I have no idea what a "data tower" is. Furthermore, in the Flock Safety FAQ included in the backup, on page 3 paragraph 7, it states that the footage is "stored in the cloud." These two statements are irreconcilable. Either it is Flock's data center or it is not. If it is not, then who is Flock's cloud service provider? Does Flock have control over those servers? What are the physical security provisions at the cloud provider's data center? These are questions that need to be answered about any storage of sensitive data.

In the section on procedure, IV(F), "F Officers will not take any police action that restricts the freedom of any individual based solely on an ALPR alert unless it has been validated." It's not clear from context what the standard for "validation" is here.

The section continues, "Officers are reminded that they need to have reasonable suspicion and/or probable cause to make an enforcement stop of any vehicle...[...]...Officers should attempt to visually match the driver or should have another legal basis for making the stop." I am very uncomfortable with the use of the word "should" here; the Fourth Amendment requires a "must." If you have not committed a crime in an officer's presence or the officer has no reasonable and articulable suspicion to justify such action, there *must* be no seizure. See Terry v. Ohio, 392 US 1(1968)5(b).

For these reasons, which I am happy to reiterate in person should you choose to reconsider this at a future meeting, I respectfully request that the Council rescind their approval.

Thank you for your time and consideration, and for all you do for our Town.

Best regards.

Edited 2:53pm 6/20/22 to correct the number of cars crossing the bridge each month; the math mistake was mine.

Edit 6/25. Followup here: http://harddeadlines.com/2022/06/23/followup-portsmouth-town-council-license-plate-readers-mt-hope-bridge