THOUGHT LEADERSHIP COMPENDIUM
February 22, 2024
SPONSORED BY
CAN AI REPLACE SHERLOCK?
Law Enforcement in the Age of AI
ABSTRACT
This paper discusses AI capabilities in deductive reasoning, problem-solving, and crime analysis. Exploring potential implications and challenges, it delves into the evolving landscape of law enforcement in the age of AI, addressing the benefits to law enforcement of teaming human intuition and machine intelligence.
Mulcare, Shweta and Singh, Vinay Professional Services Council, February 2024
LAW ENFORCEMENT IN THE AGE OF AI
Introduction
Sherlock Holmes could make extraordinary inferences from seemingly unremarkable data. But when explained, the conclusions appeared obvious. Advances in Artificial Intelligence (AI) and Machine Learning (ML) similarly possess uncanny powers of inference, with the added benefit of seeking patterns in petabytes of datai1
AI Adoption and Acceptance has Accelerated
In the last decade, AI has advanced in ways that are quite astonishing2 . Large Language Models like GPT-4 can respond meaningfully to queries much like a human conversationalist. They can write essays, compose poems, create code, produce research papers and have nuanced conversations. AI can create images from text including art, realistic human faces and scenes that never existed. Deepfake technology is blurring the lines between the real and the imagined. Add to these advances in autonomous vehicles, real-time voice translation, emotion recognition, game playing, personalized education, and we see the potential for AI to touch upon and potentially improve almost every facet of our lives. The use of AI/ML today is limited only by our imagination and there are AI tools to supplement that as well.
Keeping Pace with Data and Offenders
The need for AI/ML in law enforcement is undeniable. These tools can process vast datasets, recognize patterns, and make predictions far more reliably and quickly than humans. As data volumes and complexity increase and as the technical sophistication of lawbreakers improves; law enforcement needs commensurate investments to employ breakthrough technology for proactive mitigation and enforcement. Techniques like facial recognition and biometric screening can quickly identify people and suspects in real time and aid in investigations. The application of these technologies has led to measurable reduction in response times, improved public safety3, and given law enforcement officers tools to counter criminals. These tools are being used today to aid in proactively identifying and preventing drug movements4, human trafficking5, and money laundering.
Law Enforcement at the Speed of Relevance
AI applications can enhance just about every field of law enforcement 6 The United States leads the world in getting disparate systems to work with each, but there is much more to be done. How can AI do better? By training algorithms on more representative and new data, we could teach machines to unravel patterns and spot infractions in time to prevent them For example, identifying potential criminal hotspots and allocating officers accordingly. AI trained on completeii and cleaniii data are better equipped to aid human analysts.
i AI inferences aren’t always as elementary, which is why explainable AI is needed to increase trustworthiness in AI recommendation and solutions. See Section on Responsible AI
ii No redaction of sensitive data; fused data sets that provide a more complete picture, for example crime statistics joined to missing persons reports and gun sales
iii Represents the target population; disinformation due to contamination by adversaries
LAW ENFORCEMENT IN THE AGE OF AI
AI in Law Enforcement
The effective use of AI in law enforcement relies on data access, applying algorithms at the edge for realtime decisions and an approach that balances utility with ethical use.
AI is like Fine Wine, and it’s all About the Grapes
We need AI to extract value from data. Supervised AI systems learn patterns and make decisions based on data available for training. Without sufficient, diverse, clean, and balanced data, the systems cannot learn effectively. Data is used not just for training, but also for testing and validating AI models. This ensures the models perform as expected on new unseen data and continue to provide value even from unexpected data. The ability to continuously monitor and learn from feedback is necessary to stay relevant and useful. AI systems learn from their mistakes and are designed to continually improve and adapt to changing conditions and new information. When investing in AI, it is important to remember that the value on Day 1000 will far exceed the value on Day 1.
Secure Data Enables Effective AI
We need good data to extract value from AI. Operational AI models are only as good as the data upon which they run. We must ensure key data elements are accessible to the AI – a potential challenge for sensitive and proprietary datasets. Most of the data that law enforcement uses is sensitive and needs to be guarded. Even data that is not “sensitive” needs to be guarded against poisoning by adversaries that may be trying to influence the ML algorithms. The advent of zero trust architecture and attribute-based access control has created new opportunities for delivering value from protected data and to harness the power of AI.
Law Enforcement Applications in the Age of AI
AI offers a range of applications for law enforcement agencies to enhance their capabilities, improve efficiencies and support efforts to maintain public safety7 . Several AI applications are set to transform law enforcement over the next decade through enhanced deductive reasoning, problem-solving and crime analysis. This is what the future of law enforcement looks like:
Deductive Reasoning
• Identity Protection: AI is used in several ways for identity protection. AI algorithms are used to monitor and analyze transactions in real time to detect unusual patterns or anomalies that may indicate identity theft or fraud. AI is used for biometric authentication and analysis of behavioral biometrics. It can detect and thwart phishing attempts, monitor social media, verify documents etc.
• Real Time Policing: AI is used to predict crime hotspots and trends and enable proactive action. AI powered drones and vehicles provide additional surveillance of events and allow officers to react in an optimal manner.
Problem-Solving
• Identity Recognition: facial recognition technology is used to identify and track suspects or missing persons. The effectiveness of these algorithms improves dramatically with synthetic data. Synthetic data can be used to generate thousands of versions of an image with different eye color, hair styles
LAW ENFORCEMENT IN THE AGE OF AI
and skin tones. These synthetic training datasets in turn vastly improve the performance of facial recognition algorithms. In addition, biometric markers such as voice recognition, iris scans (unique pattern capture), retinal scans (blood vessel patterns), hand geometry (hand shape and size analysis), keystroke dynamics (typing patterns), and gait analysis are used for identity recognition.
• Crime Scene Reconstruction: AI is used in a variety of ways to reconstruct crime scenes. It is used to create 3D scene reconstruction from images and videos to aid investigators. Further, it can enhance surveillance footage, bullet trajectory analysis, blood spatter analysis, DNA analysis and a host of other tasks to aid human investigators.
Crime Analysis
• Surveillance Simulations: AI can simulate movement and behavior of crowds in public places and sports arenas. These help in training and planning for various contingencies. AI anticipates traffic and pedestrian flow patterns and identifies possible congestion points for proactive mitigations.
• Forensic Art: AI assists forensic artists in reconstructing faces. It can aid in the initial stages of sketching and thereby speed up the process. It can also be used to simulate how appearance may change over time and suggest variations to supplement gaps in witness recall.
In addition to the above listed activities, AI is currently being used by law enforcement in several ways Besides aiding in active law enforcement work, AI can also make the administrative tasks easier by automating report writing and documentation requirements. This allows law enforcement officers to focus on the more critical tasks such as context based sensemaking.
Responsible AI
It is important to balance AI advancements with ethical and privacy considerations.
Algorithmic Bias8
A primary issue is the potential for inherent biases in AI systems, which if not addressed can lead to unfair and discriminatory outcomes. Examples of these abound in several domains and the need for vigilance in ever present. Most of these technologies have been trained on large datasets and if the datasets contain historical biases or are not representative of the diverse population on which it will be used, AI could perpetuate and even exacerbate these biases9
Privacy and Transparency
Another concern centers around privacy. Surveillance and data analysis involves the collection and processing of personal data, often without explicit consent. This raises questions of how to balance public safety with individual privacy. Further, with agencies collecting and storing petabytes of data, there is risk of data breaches that could reveal sensitive information. Policies around data storage and retention will need to be formulated for the ever-increasing volumes and complexity of data. However, it may not always be possible to provide blanket policies around the proper use of data. For example, transparency may detract from security, such as with financial data. Implementing trustworthy AI is a nuanced business that must account for the particulars of a given use case.
LAW ENFORCEMENT IN THE AGE OF AI
Explainability
Most AI algorithms today use deep learning techniques. While the results are accurate, they are hard if not impossible to explain. This lack of transparency is a problem as it is hard to understand and defend some decisions or to identify errors in the decision-making processes. Government, academia and industry are working hard to enable human users to understand and trust in these outputs by being transparent in how the models are trained, tested and monitored10. This is necessary to ensure that the output from machine learning algorithms meet the “Daubert Standards”11 for evidence.
These concerns emphasize the need for ethical guidelines, transparent algorithms, and a legal framework to ensure that AI and ML technologies are used responsibly. Many government organizations have internal guidelines to avoid these pitfalls, but the need for continued vigilance cannot be overemphasized.
Conclusion
The introduction of AI in law enforcement is transformative, but also necessary to stay ahead of criminals and adversaries. The deployment of AI in the field must, however, be navigated carefully: balancing the significant advantages against potential concerns to ensure ethical and responsible actions.
An additional consideration and potential barrier to widespread adoption is cost. Implementing AI technology requires significant investment in infrastructure and training of the workforce that may be unaffordable today for some law enforcement agencies.
AI’s role in predictive policing, real-time surveillance, identity protection and recognition, and other applications show its ability to improve and aid humans in law enforcement. The use of AI in administrative tasks allows officers to focus more on the crucial aspects of law enforcement.
Despite the undeniable benefits of AI, there are valid concerns regarding privacy, potential biases, and ethical use of this technology. It is essential to address these concerns through a strict regulatory framework, continuous oversight, and a focus on training personnel in the ethical use of AI. Doing so will mitigate these concerns and ensure safe and responsible use of this powerful technology.
REFERENCES
1 Using Artificial Intelligence to Address Criminal Justice Needs
2 Artificial Intelligence’s Use and Rapid Growth Highlight Its Possibilities and Perils
3 How Assistive AI Can Be A Force Multiplier In Public Safety
4 Customs and Border Protection is using AI to crack down on fentanyl trafficking
5 Artificial intelligence and the fight against human trafficking
6 Artificial Intelligence Applications in Law Enforcement: An Overview of Artificial Intelligence Applications and Considerations for State and Local Law Enforcement
7 Artificial Intelligence, Predictive Policing, and Risk Assessment for Law Enforcement
8 Algorithms in Policing: An Investigative Packet
9 Ethical Framework Aims to Reduce Bias in Data-Driven Policing
10 What is explainable AI?
11 Machine Learning Evidence: Admissibility and Weight
Converting Cold Calls into Business Relationships
By Andrea C. McCarthy, President HARP
Oftentimes Federal Contractors want to leverage their team’s “rolodex” of Government contacts as a means of creating connections to market the firm to the Federal Government. But, what if many of the people the company wants to meet with are new to their team?
This is where “Cold Call Conversion” comes into play. Cold Calls are the front line in conducting initial outreach to Contracting Officers, Program personnel, the Chief Information Officer (CIO) etc. for the purpose of initiating a business relationship.
A person able to convert a cold contact – someone they have never met before – into a trusted relationship is a valuable asset. Who excels at Cold Call conversions? Someone who is emotionally intelligent (e.g. skilled at reading social cues etc.), confident, and has strong communication skills.
Here are some ideas on how to get started
Have a Plan: Be clear on why you are reaching out and what your desired outcomes are. Think about your long-term goals and objectives for this relationship.
Do Your Research: Government leaders continually implore Contractors to do their homework. Be prepared before doing your outreach. Know the agency mission and organizational priorities. Read the agency’s Strategy Plan and Fiscal Year Funding Request. (You will be more prepared for your next interaction.)
Develop Your Talking Points: Strategize on what you want to say and what you want to ask for. You might even want to jot it down. Being well prepared and methodical makes the process less stressful.
Generate Informed Questions: Develop questions that demonstrate you did your homework and that drive to your desired outcomes.
Make the Cold Call: You can make a “Cold Call” via an actual phone call, email, LinkedIn messaging or in-person. Build out your LinkedIn network each day by “connecting” every time you meet someone new. People feel better when they see you both have mutual connections.
Execute the Call: 1) Be brief and straightforward in your communication. 2) Present yourself as professional and knowledgeable. 3) Be personable but not overly familiar. You want the interaction to be comfortable for them and for you. Do not be pushy or overbearing. It quite possible your outreach may not work so never leave a bad impression. Washington is a small town. You are likely to bump into this person at some point.
That initial outreach is just the first step in the long and rewarding process of rapport building. (A topic for another day!) This information might seem quite obvious. But executing an effective Cold Call that yields results (e.g. the invitation by the caller for a follow on meeting) is where the challenge lies. Consider trying this approach to expand your government network and to build your business and company brand. It may seem intimidating at first, but the benefits are worth it.
Protecting the Future
The Impact of AI on Federal Law Enforcement
Sponsored by
Protecting the Future: The Impact of AI on Federal Law Enforcement
AI has taken the world by storm.
We’re already seeing its impact in nearly every industry, especially finance, healthcare, and law.
It’s taken many by surprise. In 2021, the Software National agenda predicted that artificial intelligence (AI) would be a copilot to do things like help author software within 15 years. Instead, it happened in just a few months. Federal law enforcement is likewise rapidly integrating AI technology into its practices, which provides major implications for efficiency, crime analysis, and civil liberties.
But its implementation is complicated. In October, President Biden issued an Executive Order on the safe, secure, and trustworthy development and use of AI. As such, harnessing the power of
AI must be put in the hands of experts who can deploy the technology with precision and security in mind.
To ensure that government agencies and law enforcement continue to provide optimal services to the public, they must continually explore AI-driven solutions.
AI history and background
While the launch of ChatGPT in November 2021 brought AI to the forefront of international discussions, its use in law enforcement is not new.
The FBI began pilot programs to adopt basic predictive analysis for hot spot mapping of crime locations in major cities—as well as forms of facial recognition software—as early as the 1990s. It found its way into data processing too, like in 2001, when
machine learning (ML) assisted in the Enron investigation.
Since then, predictive policing has been on the rise. Accenture reported that 76% of police departments use AI and ML tools for video analysis, and Capgemini found that roughly 40% of agencies use AI to combat cybercrime, fraud, and theft.
$3.62 Billion projected budget for AI in law enforcement by 2025
The future of AI in law enforcement
EM360 reported the use of AI in law enforcement is expected to grow at an annual rate of 19.6% through 2025. By
then, that industry’s projected budget for AI tools will be $3.62 billion. And with good reason. Mordor Intelligence reports AI technology could potentially reduce the workload for investigators by 30% just by implementing it in facial recognition practices alone. A UC Berkeley study found that AI integration in Shreveport, Louisiana resulted in a 35% reduction in burglaries and a 26% reduction in
vehicle thefts.
Here are some ways SOSi predicts AI will continue to impact law enforcement practices:
Enhancement of predictive policing
Machine learning algorithms can analyze historical crime data to identify patterns and trends to forecast crime hotspots and potential criminal activity. For example, the Santa Cruz Police Department in California uses an ML
SOSi CEO Julian Setian on how SOSi is implementing AI and ML tech
Emerging technologies are almost all driven by the need to make faster, more informed, and more trusted decisions. At SOSi, we’re using AI and ML to help the government make decisions faster.
This is done by providing real-time analysis and insights on millions of data points daily.
Security and trust is paramount when implementing these new technologies, and blockchain-based security provides an exciting approach for establishing and maintaining trust across decentralized systems. Blockchain’s innate auditability and encryption can improve security and accountability in government operations
involving identity management, benefits distribution, record-keeping, and a host of other areas.
But to fully realize the benefits of these emerging technologies: sustained investment is needed for updating legacy systems, training civil servants and developing supportive policies.
The key is integrating AI, blockchain, and other innovations natively into government infrastructure workflows.
Technology can be transformational, but human oversight and ethics have to remain central to its implementation.
system to forecast where burglaries are likely to occur. The system analyzes data on past burglaries, weather patterns, and other factors to identify areas with a high risk of burglary. Since implementing the system, the department has seen a 19% reduction in burglaries in the forecasted areas.
Investigations and intelligence
AI tools can scan massive sets of data from various sources to uncover connections in cases that humans may have missed. This can generate new leads and evidence.
Image recognition tech can also be used to match faces and license plates to databases.
Surveillance
The increasing use of AI in analyzing footage from body cameras, street cameras, and drones can enhance officers’ situational awareness and ability to respond better.
SOSi Director of Law Enforcement Services Mark Nemier said he believes AI can do this by improving agencies’ ability to process the large-scale processing of lawfully intercepted voice communications.
“The most immediate impact AI will have in law enforcement will be in the area of transcription and translation,” Nemier said. “If you have a recorded conversation, you can use the technology to process it almost instantly. Then, you can use human intervention for quality control.”
Nemier also said he predicts that as AI tech improves over time, transcription and translation processing could also be applied to Title III surveillance as well—assuming all legal hurdles are cleared and there’s a cultural acceptance of the AI-assisted evidence among investigators and prosecutors.
Analytics and authentication
Data mining and analysis of incident reports, crime stats, and resource allocation can help identify inefficiencies and inform better strategies.
AI and ML can also help authorities authenticate information faster.
SOSi VP of Civil Solutions Charles O’Brien said, for example, that AI could greatly improve the processing of newly arrived non-citizens at the U.S. border.
3 Ways AI and ML are Impacting SOSi Customer Missions:
1 – AI-powered Cyber Defense
We deliver AI and ML-enabled cyber capabilities to key customer missions such as defending warfighting U.S. coalition data systems in the global Mission Partner Environment
2 – SOSi’s subsidiary Exovera’s ExoINISIGHT
A platform that unleashes generative AI and ML to provide organizations with a near real-time
“As we’re trying to process noncitizens, they’re presenting all kinds of documents to confirm their identity,” O’Brien said. “But how do we know they are who they say they are? By implementing AI into our processing systems, our point-of-contact communications could be almost instant, which will not only improve efficiency but also limit our country’s risk exposure.”
Training and retention of institutional knowledge
SOSi CTO Kyle Fox on how corporate experimentation furthers AI progress
Technology is advancing so fast, and it is important to understand its limitations and how to use it effectively. Many companies lack a culture of experimentation, which can hinder progress. It is crucial for everyone to understand that experimentation is necessary for success in any role. Building a diverse team of creative problem solvers, including IT professionals with various backgrounds and skill sets, can lead to a stronger team. Compliance experts and out-of-the-box thinkers can help drive innovation. By working together, the team can achieve optimal results.
picture of key patterns and developments within the foreign information environment.
3 – Leverage AI to support operations at the border
From language translation to providing access to medical services, our AI-powered solutions help support agents at the border deliver needed services to migrants.
One area that AI will be especially helpful to law enforcement agencies is training and knowledge retention. The percentage of agents who are eligible for retirement has increased over the last several years, and though agencies such as DHS have doubled down on recruiting efforts to maintain workforce quotas, efficient and effective training will become a critical component to their success.
AI simulations and virtual environments provide alternative methods for training law enforcement to make quick decisions in complex situations, meaning that they’ll be able to reach peak performance in a quicker timeframe.
Likewise, AI’s predictive and patternidentifying capabilities can also help to retain institutional knowledge within agencies—even as the workforce turns over.
“I’ve seen Border Patrol agents who are amazing at catching narcotics because they’ve spent 25 years patrolling one spot,” SOSi Operations Manager Ramiro Garza said. “They know when a sensor hits in a certain way, there’s a high probability there’s going to be a narcotics load in that area. But once the agent retires, that knowledge retires with them, and it will take time for a
new agent to learn the same tricks.”
Garza said that integrating more ML tech will ensure such institutional
SOSi Senior Cyber Security Engineer Sean Paul
on how AI will impact software development
AI is rapidly changing the way we develop software. AI-automated tools are beginning to take over many of the tasks that were once done by human programmers—especially the tedious ones.
Letting AI do the repetitive tasks results in freeing up human programmers to focus on more creative and strategic work.
AI is also having a major impact in the field of software development through code generation. You can now use AI to generate code from natural language descriptions.
This makes it easier to establish the early development of code to standards. It can also be dynamic to the programmer’s needs.
knowledge will stay with the agency. Sensor activity can be instantly crossreferenced with narcotics capture statistics and deliver any agent the probability of a narcotics load being present.
Trusting AI
A key focus of taking next steps in AI implementation will be trust on many levels.
Security is paramount to safe adoption. SOSi Chief Technology Officer Kyle Fox said applying techniques such as the Zero Trust security framework— which assumes all users, devices, and network traffic are untrusted and require continuous authentication and validation—is a promising approach.
Another consideration is ensuring that law enforcement officials can rely on AI-generated results. Concerns
There’s also bug hunting. AI could identify bugs in code, comparing it to good examples. They can also fix bugs automatically, so when you’re using it as a part of the automatic code testing process, it can identify and eliminate bugs early in development.
Let’s not forget security. Some AI models are programmed to detect the tech security vulnerabilities in code. They can also generate code that is more secure by default.
While the impact of AI on coding is still unfolding, AI is certainly here to stay. It makes coding easier, faster, and more secure. Believe it or not, this is good news for both the developers and users.
to transfer funds into illicit bank accounts. “We’ll have to develop countermeasures to combat red herrings and synthetic media.”
The first step in solidifying that trust is for agencies to work with developers who understand their unique difficulties and complexities of their processes.
“There are many tech companies who can develop AI solutions but not many who have had such a long-term, multilayered relationship with federal law enforcement agencies,” Fox said. “Working with companies like SOSi allows for better transparency and collaboration when building solutions, which guarantees the tools will make a difference in the lives of those who are using it.”
about the accuracy and fairness of AI algorithms—as well as the potential for bias in the information used to train AI machines—must be addressed by continually training and testing them on large and diverse data sets and implementing quality control measures to detect and correct errors.
It’s also important to remember that as law enforcement develops its AI capabilities, so do lawbreakers.
“We’re already seeing criminals employ deepfake tactics to perpetrate crimes and spread misinformation,” SOSi Capture Director Ivan Veskov said, pointing to several examples where cybercriminals used AI-generated deepfakes to impersonate company executives and direct employees
SOSi’s core mission is to promote and protect the interests of the U.S. and its allies around the world.
Since our founding in 1989, we have empowered our employees to develop solutions that break through barriers, inspire innovation, and build resiliency. Today, our motto of “Challenge Accepted” resonates through our work modernizing and securing legacy government IT systems, driving innovation for the U.S. Department of Defense and Intelligence Community, managing critical government facilities and infrastructure, delivering critical intelligence analysis, and supporting enforcement, humanitarian, and asylum operations at the border.
Yet, what sets SOSi apart is not what we do, but who we are. As a privately held and founder-led company, our creative and spontaneous culture enables us to be bold, act fast, own and take responsibility for our results, and build and maintain relationships that matter. SOSi offers a large company’s depth, breadth, and infrastructure, and the missionfocused agility and innovation of a small business.