AI’s extension boundaries and its required data
Artificial intelligence is rapidly changing the way we live, work and manage. In public health and public services, AI tools promise increased efficiency and faster decision-making. But below the surface of this transformation is a growing imbalance: our ability to collect data exceeds our ability to control it responsibly.
This is just a technical challenge, which is to become a privacy crisis. From predictive policing software to surveillance tools and automated license plate readers, data about individuals is accumulating, analyzing and acting at an unprecedented rate. However, most citizens don’t know who owns their data, how it is used or whether it is protected.
I’ve seen this already. As a special agent for the FBI network and now CEO of a leading public safety technology company, I have worked in the government and the private sector. One thing is clear: if we don’t solve the current way of handling data privacy, AI will only make existing problems worse. There is one of the biggest problems? Walled garden.
What is a walled garden and why is it dangerous in public safety?
The Walled Garden is a closed system in which a company controls access, traffic and usage of data. They are common in advertising and social media (think platforms Facebook, Google, and Amazon), but are increasingly appearing in public safety.
Public safety companies play a key role in modern policing infrastructure, but the proprietary nature of some of these systems means they are not always designed to interact with other vendors’ tools.
These walled gardens may offer powerful features such as cloud-based body video or automated license plate readers, but they also create a monopoly on how data is stored, accessed and analyzed. Law enforcement agencies often find themselves in long-term contracts with proprietary systems that do not talk to each other. result? Dispersed, isolated insights and effective responses in the community, where most importantly.
The public doesn’t know, this is a problem
Most people don’t realize how much of their personal information is flowing into these systems. In many cities, your location, vehicle, online activities and even emotional state can be inferred and tracked through AI-powered tools. These tools can be sold as an escalation to combat crime, but are easily abused without transparency and regulation.
Not only does data exist, but also in the walled ecosystems that are controlled by private companies with minimal supervision. For example, tools such as license plate readers are now in thousands of communities in the United States, collecting data and feeding it into their proprietary networks. Police departments usually don’t even own the hardware, they rent it, meaning data pipelines, analytics and alerts are determined by the vendor, not by the public consensus.
Why should this cause red flags
Artificial intelligence requires data to run. However, when the data is locked within the walled garden, it cannot be cross-referenced, validated, or challenged. This means that decisions about who is removed, where the resources go, or who is marked as a threat are made based on partial (sometimes inaccurate information).
risk? Improper decisions, potential civil liberties violations, and the growing gap between the police department and the communities they serve. Transparency erosion. Trust evaporates. And innovation is stifled because it cannot enter the market unless new tools meet the constraints of these fence systems.
In one case, the license plate recognition system incorrectly tags a stolen vehicle based on outdated or shared data without verifying information about how the decision was made across platforms or audits, and officials can take false alarms. We have seen incidents of flawed technology leading to false arrests or escalation of confrontation. These results are not hypothetical, they occur in communities across the country.
What do law enforcement officers actually need
We need an open ecosystem, rather than locking data, to support secure, standardized and interoperable data sharing. This does not mean sacrificing privacy. Instead, this is the only way to ensure privacy protection is performed.
Some platforms are working to achieve this. For example, FirstTwo provides a real-time situational awareness tool that emphasizes responsible integration of publicly available data. Others, such as Forcemetrics, focus on combining different datasets such as 911 calls, behavioral health records and previous event history to give officers a better background on the scene. But it is crucial that these systems are built in public safety needs and community respect, not afterwards.
Establish privacy-first infrastructure
A privacy-first approach means more than just editing sensitive information. This means restricting access to the data unless there is a clear legal requirement. This means documenting how decisions are made and third-party audits are enabled. This means working with community stakeholders and civil rights groups to shape policy and implementation. These steps lead to enhanced security and overall legality.
Despite technological advancements, we are still operating in a legal vacuum. The United States lacks comprehensive federal data privacy legislation, allowing agents and suppliers to set rules along with the rules. Europe has GDPR, which provides a roadmap for consent-based data use and accountability. In contrast, the US national policies are pieced together in pieces that cannot adequately address the complexity of AI in public systems.
That needs to be changed. We need clear, enforceable standards around how law enforcement and public safety organizations collect, store and share data. We need to include community stakeholders in the conversation. From procurement to implementation to daily use, consent, transparency and accountability must be baked into every level of the system.
Bottom line: Without interoperability, privacy suffers
In terms of public safety, life is online. The idea that a vendor can control access to mission-critical data and limit how and when to use it is not only inefficient. This is immoral.
We need to transcend the myth of innovation and privacy conflict. The AI of the person in charge means a more fair, effective and responsible system. This means rejecting vendor lock-in, prioritizing interoperability and requiring open standards. Because in a democracy, no company should control the data that decides who gets help, who stops or is abandoned.
