The definitive guide to using AI in policing

Axon is committed to leveraging AI innovation ethically to revolutionize community safety while prioritizing the mitigation of biases and risks. This approach ensures that AI serves as a force multiplier, enhancing the capabilities of public safety professionals while preserving the critical role of human decision-making. Simply put, AI should augment operations, never replace the human.

In this article, we will explore several key considerations for law enforcement agencies who want to leverage AI in policing in a responsible way. Let’s get to it.

Responsible AI Development: A Methodical Approach

As a leader in the public safety space, Axon has carefully developed a methodology that equally considers ethics and efficacy when it comes to leveraging artificial intelligence. Axon’s Ethics-by-Design methodology guides the responsible development of AI solutions:

• Problem-Centric Approach: We always start with a problem that needs solving when we invest resources into developing AI technology, and we prioritize non-AI solutions when those prove effective.

• External Collaboration: Engaging diverse external experts ensures fair and equitable AI development.

• Risk Assessment and Mitigation: Risks are identified and mitigated throughout the design process, prioritizing user safety and fairness.

• Diverse Data Training and Evaluation: AI models are trained and evaluated with diverse data to ensure performance across various scenarios and demographics.

• Human Oversight: Safeguards are built into AI technology, ensuring human decision-making in critical moments.

• Continuous Improvement: Rigorous testing and ongoing monitoring post-release ensure fairness and accuracy.

Learn more about our Responsible approach to AI development.

Civilian vs Public Safety AI Tools

Axon designs our AI solutions with public safety in mind, building secure, efficient, and purpose built solutions.

Draft One is a great example when comparing the qualities of a public safety grade solution with a consumer grade one.
Key distinctions between our solutions and consumer-level generative AI include:

• Data Security: Your data is completely secure within the Axon network. On the occasions where we need to test with real customer data, we request permission to enroll them in our voluntary, privacy-centric program — all along working within the confines of our data sharing agreement.

• Hallucination Reduction: The underlying model we used for transcription is Open AI’s GPT-4 Turbo, and we calibrated the model to prevent speculation or embellishments. Draft One reports stick to the facts and require officers to review and add any missing pieces of information.

• Safeguards: Draft One requires users to insert info that is missing. An officer cannot copy and paste the text until they have reviewed and updated every insert statement. Once all the insert statements are updated, officers must sign off on the narrative’s accuracy and attest to their ownership of the report before submission.

• Tailored for Law Enforcement: Draft One is a closed system with industry-specific safeguards, trained to write great police reports using the best resources available.

• Integrated with the Axon Ecosystem: As soon as an officer stops recording with their Axon Body 3 or Axon Body 4 camera, the audio from the recording begins loading to the cloud via LTE.

Learn more about the differences between Draft One and consumer grade generative AI tools.

Current Applications of AI in Policing

AI is already integral to various aspects of policing. For instance, in the following video, Lafayette PD share how solutions like Draft One are saving officers hours and alleviating burnout:

Here are a few other ways AI is being leveraged in law enforcement today:

• Computer Vision: Enabling functions like redaction, evidence search, and automatic license plate recognition.

• Speech Recognition: Facilitating auto-transcription and body-worn camera (BWC) video audits.

• Speech to Text: Assisting officers with report writing through dictation.

• Translation: Enhancing communication by translating languages.

• Generative AI: Streamlining report drafting processes through summarization.

Axon has a growing list of solutions that leverage AI in some way. Learn more about these solutions in the article, how AI is shaping public safety in a responsible way.

Responsible Use of AI in Policing

We recognize the immense promise of ethical AI innovation. Our mission is to harness cutting-edge AI technology to revolutionize community safety, all the while prioritizing the rigorous mitigation of biases and other potential risks.

One of our methods for building responsible AI solutions is to implement safeguards that keep people as the decision-makers, seeking to force multiple their efforts rather than to replace them. We also work with police to help ensure they follow best practices.

As you consider your AI policies, it is important to consider the following things:

• Always ensure the user is in the loop: AI should not replace human decision making, so officers need to be trained and reminded that they are responsible for making the final decisions.

• Remember that AI is an assistant: There are many who fear that AI will replace jobs, but we believe that humans work best when they partner with technology. AI cannot replace humans, but it can assist them. You need to understand the strengths and weaknesses of AI, and make sure you leverage its strengths.

• Data Protection: Not all AI solutions are the same when it comes to security. As an agency, ensure your people never share sensitive data with data-hungry AI tools, and instead select tools with proven, public safety grade security.

• Evidence Integrity: The Justice system relies on truth, so AI in police work should always stay focused on the facts. Agencies need to prioritize the importance of the chain of custody when it comes to digital evidence, and AI shouldn’t elaborate or alter evidence.

• Consider Alternatives: Agencies should explore non-AI solutions where appropriate - sometimes AI provides the best solution, but other times it is not required.

• Remember AI is not perfect, but it can help reduce naturally occurring human errors: AI is not perfect, but humans miss things all the time, too. Leveraging AI while keeping humans in the loop can help reduce the number of errors people make on their own.

To learn more about Axon’s values-based approach, read our framework that guides our innovation.

Looking Ahead: Shaping the Future of Policing

As technology continues to evolve, the future holds immense promise for responsible AI in policing. New solutions will emerge, each presenting opportunities for innovation and advancement.

At Axon, we remain steadfast in our commitment to responsible AI development, driving forward initiatives that enhance intervention efficiency while preserving the fundamental role of human judgment.