"How is your business structurally different from the data collection companies like a Flock or an Axon?" the host asked, part way into a Training Data conversation with Nick Noone and Ben Rudolph, the founders of Peregrine.
Noone's answer was a description of the industry he entered. Peregrine has been working in public safety since 2018, he said, and most of the companies that came before it built their businesses "fundamentally on the back of data collection" — a sensor or a form installed inside a customer's operation, with the product being the input and the storage of that input. Once a company holds a customer's information and has a foot in the door, he said, it can sell more things, and the additional things tend to be more collection systems.
"Peregrine is like the inversion of that entire model," he said. The company's business, as he described it, is joining information that already exists in an agency's separate systems so that people can get more precise answers from it. It sits on top of what is already there rather than adding new streams of its own.
What is already there
The records in question are scattered by default. In a January 2023 company article, Tim Shriver describes police information spread across body-worn cameras, automated license-plate readers, dispatch systems and records-management software, with the same person appearing in one system with a middle initial and in another without it. Peregrine's proposed role there is to integrate and organize those existing records so a department can search across them, build reports and answer public-records requests without assembling everything by hand.
Peregrine's customers are no longer only police departments, Noone said: police, fire, emergency management and health services are all in the mix now, with what he called the through line of "delivering safety and prosperity to cities."
The anti-network-effect proposition
The usual advantage in this market is a network effect: the more data a system holds, and the more customers who feed into it, the more valuable and harder to leave it becomes. Noone said the world is already worried about that — "the amalgamation of data" inside some central body, public or private, and what it might do with privileged information. He wants the opposite. Peregrine, he said, aims not to be "in any shape or form in the business of bringing more data to the customer that they don't already have." The problem he says he is solving is that agencies cannot use the data they have, or cannot use it in a way that is secure and trusted by the communities they serve.
He also described the pull in the other direction. An organization improvising its way to something useful could reach for open-source data it may or may not be permitted to touch under the laws, regulations, ordinances and standard operating procedures that govern it — a "gray zone," in his words. Nobody should break those rules to get AI into a complicated environment, he said; the job is to protect the data and the institution instead. Because these organizations sit on top of what he called astounding network effects, refusing to exploit them is, as he put it, "almost like the anti-network effect proposition": preserve each agency's ownership and ways of working, and build the connective tissue between them separately. He described the company as "a big practice in letting go" — building transparent solutions, making them transparent to customers and their constituents, and not trying to grow too fast.
Who owns it, and who may look
Rudolph put the ownership principle plainly. Each institution owns its own data; the organization serves a community, so the records are the community's and the organization's, "not Peregrine's data." From there the question becomes control: the ability to roll a system out securely inside a department and share selected pieces of information outward only when needed.
His comparison for the alternative was physical. Without tools like this, he said, "you will put a bunch of data in the back of a car and drive it across a city, and there's actually less control there." A hard drive handed to a neighboring agency carries no rule about which rows may be read or how long they may be kept. The controls he wants are the ones that let a department share without oversharing and trust what it has handed over.
Peregrine's undated trust documentation sets out what the company says those controls are. Permissions can be set by role or attribute down to individual rows and fields. Access to sensitive data can require a purpose-specific justification, and searching or exporting can be restricted; actions are recorded with user identities and timestamps. Data lineage tracks how information was ingested and transformed, and provenance identifies where it came from. Customers set retention and deletion rules, and the company says it does not sell customer information, repurpose it, or train AI models on it. Sharing stays subject to the data owner's access rules, and the company says interoperability is meant to keep information portable to other vendor partners. For AI workflows, the documentation says a customer must enable them, that they respect the same access boundaries, that outputs link to the evidence behind them, and that critical decisions stay with people.
Those commitments are not unusual as claims. Flock, the license-plate camera company named in the host's question, describes its own arrangement in similar terms: customers own their data and must authorize sharing, searches require a documented case-related reason, audit trails are kept, and plate observations are deleted after 30 days by default or as retention law requires. Those are stated product rules on both sides. The argument Noone is making is about business structure — what a company is paid to grow — rather than about a feature list no competitor offers.
The line the company will not draw
The host asked where morally difficult technology decisions land, using facial recognition as the example. Noone's answer was that a Silicon Valley company imposing a general-purpose decision on an industry is "fundamentally wrong." Asserting whether a technology may be used, or what a retention policy should be, is "very much not how we think." He said most public-safety agencies in America and their communities choose not to implement facial recognition; some take a strong stand, and for others it is a matter of subjective preference. Creating "a technological red line without understanding the texture and context" is not, he said, a boundary the company believes it can assert over a customer.
"So follow the customer, follow the law," the host summarized. Noone agreed, and added the part he considers the company's work: exposing the limits and institutional norms that apply, and helping a customer see considerations it may not know about. He described how the field actually learns — "it's astounding how cities call each other" for advice — and said helping agencies get the best possible information more easily improves even their decisions about which technologies to use or not.
That is where the position bites. Earlier coverage of license-plate camera federation turned on an argument from Salim Ismail that what once protected people was friction: following everybody used to be too expensive to do, and cheap pattern matching removes that limit. Peregrine's answer leaves the decision about how far to look with each jurisdiction and its law, not with the vendor — while the company's stated purpose is to make looking across an agency's own records far more precise than it used to be.
The founders' reply to the prospect of serving many more cities was that it requires an operating model "fundamentally about infrastructure": technology that lets organizations do with their data as they would like and as they are required to do, with each jurisdiction's uniqueness preserved. The economics were offered as the reason it is possible at all. Supporting state, county and city public safety defeated a lineage of business units inside large organizations, they said, because delivering tailored solutions at a price these organizations could afford was never possible; dropping the marginal cost below a million dollars a year, by an order of magnitude or more, is what earned the company the right to try. As for the power that comes with sitting in that seat, Noone's line was: "who are we to think that we hold any power over that? I mean, it's the institution that has the power."