The AI landscape — why this matters now
Just a few short years after the introduction of AI-powered chatbots onto the market, artificial intelligence is everywhere. AI is influencing and impacting nearly every part of our lives.
In law enforcement, automation tools with AI-like characteristics have been used since the early 2000s. In the past several years, developers have created numerous AI-powered technologies that promise to transform how criminal investigations are conducted.
At a 2025 Commission on Accreditation for Law Enforcement Agencies (CALEA) Conference, 217 law enforcement attendees were asked about their AI use:
- A reported 38% said they were using AI
- Up to 20% said they were not using AI
- Another 32% said they were evaluating AI tools
There are good reasons for agencies of all sizes to be intrigued. Evidence suggests that AI-assisted investigations are already dramatically reducing the time officers spend on research and documentation. Among its many capabilities, AI can:
- Identify subjects and associates across jurisdictions using available records and data sources
- Analyze financial patterns and suspicious activities across disparate data sources
- Quickly identify patterns and trends across jurisdictions, thus potentially connecting subjects to additional cases
But here’s another important fact — not all AI tools available to law enforcement can actually help them. We’ve all encountered what is termed “consumer-grade AI” on search engines, PC laptops, and online. This form of AI works well enough for drafting basic documents and emails, summarizing online searches, creating online content, managing personal schedules, and other tasks. But it doesn’t have the reliability, the comprehensive data, and the defensibility in court required by law enforcement investigators.
What agencies require is an AI solution that has been purpose-built for investigations. Since so much depends upon case integrity and courtroom defensibility, it’s crucial that investigators, commanders, and agency technology decision-makers perform thorough due diligence when evaluating AI tools for investigative use.
Chapter Two
The 5 pillars of AI evaluation
Since not all AI tools can meet law enforcement’s stringent requirements, agency leaders need a framework that can help them evaluate the tools on the market and determine which ones can boost rather than hinder investigative effectiveness. The following five pillars offer just such a framework.
Pillar 1: Data sourcing and trusted content
It goes without saying that investigators require information that is verified and trusted. Inaccurate data can lead investigators down dead ends and wrong routes. It can not only waste time — it may jeopardize case outcomes if an investigation results in an arrest.
A major aspect of AI’s power is its ability to access and sort through vast quantities of digital data instantly, then rapidly glean verified insights from that data. But it’s crucial that law enforcement agencies be aware of how and where an AI tool obtains its data.
By and large, AI uses one or both of the following automated techniques:
Open-web scraping
Open-web scraping is the automated extraction of data from public websites. This technique collects data in formats that allow users — such as law enforcement investigators — to easily retrieve and analyze it. Web scraping can retrieve a great deal of necessary and useful information, but it does have its limits.
Data curation
This technique focuses on accessing more specialized data, including public records, such as those involving incarceration, and other premium proprietary sources across jurisdictions. “Curation” is the capability of actively managing and organizing these datasets so that they remain rigorously current. Curation can also resolve problems involving conflicting or incomplete records.
Both open-web scraping and data curation are essential for a thorough sourcing of data. Combining these methods helps ensure that the data is of the highest quality, and that investigators can trust an AI tool’s outputs — that is, the information the tool delivers based on an investigator’s query.
Pillar 2: The consumer-grade AI problem
When it comes to investigative use, consumer-grade AI exposes agencies to numerous risks, most notably:
- Security gaps
- Data retention problems
- A lack of a chain of data custody
- Compliance, privacy, and liability issues
- Uncertainty around admissibility of AI outputs in court
In short, consumer-grade AI platforms lack the level of control over their data that law enforcement requires. This insufficient control can put agencies at risk because it can open holes in the prosecution’s case that the defense can exploit.
That leads directly to the next pillar.
Pillar 3: Courtroom defensibility — hallucinations vs. citations
When law enforcement investigators use an AI tool to help prosecutors, the output the tool delivers must stand up in court. This fact is why investigators need to be aware of hallucinations and citations.
- A hallucination occurs when an AI tool generates false, misleading, or entirely fabricated information. In other words, the tool delivers fake output. There are several reasons why a tool might hallucinate. Those reasons mostly boil down to an absence of reliable data. Very simply, if the tool can’t find the information the user is requesting, it often makes it up instead.
- A citation refers to AI output that can provide verifiable sources for the information or data it delivers. Every piece of data can be traced to a reliable, authoritative source. Cited output is defensible — it can stand up in a court of law.
In the recent years, there have been numerous instances where AI hallucinations have wrecked legal cases. In most of these situations, the researchers cited legal cases that didn’t actually exist. Clearly, if an investigator’s report that contains hallucinations is submitted as an official record or used to obtain a warrant, it could corrupt the chain of evidence.
AI tools that are purpose-built for investigations can help agency investigators avoid these risks in several ways. Most notably, professional-grade tools can — or should — create audit trails. Audit trails help investigators personally confirm the reliability of data sources. By ensuring transparency, accountability, and evidence integrity, audit trails provide a huge piece of defensibility.
Pillar 4: Real-world time savings
The length of time required to investigate a crime depends upon a variety of circumstances. Simple property crimes might require only a few hours, while more complicated violent crimes may take weeks or months. Investigators know that the first 48 hours of an investigation are critical. The quicker evidence and information are found and scrutinized, the faster the crime is likely to be solved — and prosecuted.
This is why what is called AI compression can be invaluable to investigators. The standard manual approach to investigating an incident can require hours of an investigator’s time. AI automation can compress elements of that investigation from hours to minutes. This ability allows investigators to spend more of their time on other case-related tasks, such as field follow-ups and interviews.
Pillar 5: Documentation and report writing support
Like the research part of an investigation, manual report writing takes hours of time for officers and investigators. It’s not due solely to the writing itself. Gathering and incorporating the documentation required for watch commander reports, CompStat meetings, and — for high-visibility cases — briefings to city and county officials can be burdensome. There’s always the fear that the report may accidentally overlook key documentation, which can create evidentiary risk downstream. Again, both prosecutors and defense attorneys will want proof that the information in an investigative report has been fully vetted for reliability.
AI automation can simplify report and documentation processes by creating summaries of pertinent documents. But law enforcement investigators require sourced summarization. A tool should generate draft summaries with inline citations. This feature maintains evidentiary integrity throughout the documentation process so that prosecutors can back every aspect of their case with verified information.
One caveat is that investigators should review and finalize these citations so that the information and sources are accurate and true — and that any AI-generated hallucinations have been deleted.
Chapter Three
5 questions to ask AI vendors
The evaluation framework in the previous chapter can be translated into five fundamental and actionable questions that law enforcement agencies can direct to potential vendors.
What are the sources of your data?
In other words, how reliable is the data that your tool accesses and uses as a basis for its outputs? How can you prove that reliability?
Can your tool’s outputs be readily traced to those sources?
A professional-grade AI tool for investigations should be able to deliver audit trails that allow investigators to confirm the sources’ reliability.
How do you ensure law enforcement data security?
If the data isn’t protected, it can cause issues for investigations and court proceedings.
What are your documented time savings?
An AI tool that can’t prove it will save an agency money is not worth the investment.
How does your tool support report writing and documentation for court?
For instance: Are its source summations reliable and free of hallucinations?
Chapter Four
CLEAR Investigate — meeting every standard
With the stakes so high, law enforcement investigators need a thoroughly reliable AI-powered tool. Thomson Reuters CLEAR Investigate platform meets all five of these evaluation pillars.
This investigative AI platform is powered by CLEAR, a Thomson Reuters technology that rapidly analyzes a wide variety of high-quality datasets. CLEAR Investigate has a premium and open-web data sourcing that allows law enforcement investigators to quickly find missing children, locate persons of interest, apprehend criminals, and perform other critical public-safety missions. By automating hours of tedious work — summarizing lengthy indictments, cases, and a variety of internal and external files — CLEAR Investigate allows investigators to focus their time and expertise on crucial analysis and high-priority leads.
In addition, CLEAR Investigate offers the first agentic AI investigative solution designed for law enforcement. Very simply, agentic AI operates like an autonomous agent that can execute multistep processes with minimal human supervision, from start to finish. It moves beyond standard search-driven processes to include vast collections of public and proprietary records, delivering reliable summaries and defensibility through thoroughly documented audit trails.
One final due diligence best practice to note is that AI technology is continually evolving and improving. As law enforcement agencies learn more about AI, it’s essential that they determine whether a vendor’s solution has the capabilities to learn and grow along with them.
Visit us to learn more about CLEAR Investigate and how it accelerates and optimizes law enforcement investigations.
