On September 4, Mark Genatempo, Senior Fellow at Rutgers University Miller Center on Policing and Community Resilience, presented “AI Fundamentals for Public Safety.” Watch the recording here.

Your law enforcement agency just secured funding for new technology initiatives. Leadership wants to explore AI applications for daily operations. Officers have mixed reactions—some excited about efficiency gains, others worried about reliability and accountability. You know AI could help your department serve the public better, but you also know implementation failures can damage both operations and community trust.

You're facing the same tension between opportunity and risk that departments nationwide are experiencing. The key insight emerging from law enforcement AI implementations is surprisingly straightforward, said Mark Genatempo, Senior Fellow at the Rutgers University Miller Center on Policing and Community Resilience: treat AI like any other power tool in your professional toolkit.

The Power Tool Framework

 "AI is meant to support decisions, not make the decisions,” Genatempo said. “It makes it faster and easier, but the person operating or holding the tool is responsible for the work."

 "AI is meant to support decisions, not make the decisions,”

Just as you wouldn't hand a power saw to untrained personnel, you can't deploy AI without proper preparation. The tool is able to amplify your capabilities, but you remain responsible for the outcomes.

This perspective shapes your evaluation process: What training will your officers need? What oversight mechanisms will you establish? How will you maintain accountability when AI assists in decisions affecting community members?

Know What You're Actually Implementing

Before moving forward with any AI system, you need clarity on what type of tool you're considering. The "AI" label covers vastly different technologies with different capabilities, risks, and requirements.

Machine Learning systems analyze patterns in your existing data to improve performance over time. You might use these for processing evidence files, analyzing crime patterns, or streamlining report classifications. These systems require substantial historical data and ongoing monitoring to ensure they perform as expected.

Predictive Analytics tools help you forecast future needs based on historical patterns. You could apply these to patrol allocation, resource deployment during events, or identifying areas with higher service demands. However, these tools extrapolate from past patterns—they can't account for unprecedented incidents or changing crime trends.

Generative AI creates new content like text, images, or responses. You might pilot these for drafting routine reports, creating training scenarios, or generating initial policy language. These tools require careful review since they can produce plausible-sounding but inaccurate information.

Understanding these distinctions helps you match the right tool to your specific challenges and set appropriate expectations for performance and oversight.

Manage the Risks You Can't Ignore

The same characteristic that makes AI valuable—learning from patterns in data—creates your biggest implementation risks. As Genatempo noted during the workshop, "AI is only as good as the data and assumptions behind it. The old adage, bad data in, bad data out, is never more true than in artificial intelligence."

Data quality becomes critical. If your historical data reflects past disparities in enforcement patterns, AI systems trained on that data may perpetuate those disparities. Before implementation, audit your data for completeness, accuracy, and potential bias. This represents technical necessity, not just procedural compliance—it determines whether you maintain community trust.

Over-reliance threatens judgment. When AI recommendations become routine, your officers may stop questioning results. Build systematic checks into your processes. Require human review for decisions affecting individual community members. Establish clear protocols for when officers should override AI recommendations.

Lack of transparency undermines accountability. Community members deserve to understand how decisions affecting them are made. Ensure your AI systems can provide clear explanations for their recommendations. Document your decision-making process so you can explain both AI inputs and human reasoning to stakeholders.

Build Responsible Adoption Practices

Successful AI implementation requires the same systematic approach you'd use for any major operational change. Start with pilot projects that have clear success metrics and limited scope. This allows you to learn without exposing your entire operation to risk.

Establish clear policies before deployment. Define when AI assistance is appropriate, what level of human oversight is required, and how you'll handle system failures. These policies protect both your officers and the communities you serve.

Invest in ongoing education. AI capabilities evolve rapidly, and your officers need to understand both current limitations and emerging possibilities. Regular training ensures everyone can use these tools effectively while maintaining appropriate skepticism about their outputs.

Maintain community engagement. Proactively communicate about AI use in your operations. Explain what these tools do, what safeguards you've implemented, and how community members can provide feedback. Transparency builds the trust necessary for sustainable innovation.

"Courts, oversight boards, and [the] public will always hold humans, not machines, accountable,” Genatempo said. 

Your Next Steps

If your police or law enforcement agency is  considering AI adoption, start with these concrete actions:

• Conduct an internal needs assessment. Identify specific processes where AI assistance could improve efficiency or accuracy. Focus on high-volume, routine tasks where officer's expertise can focus on judgment rather than processing.

• Audit your data infrastructure. Assess data quality, completeness, and potential bias in the information you'd use to train or operate AI systems. Address data issues before considering AI implementation.

• Develop implementation standards. Establish clear criteria for AI adoption, including requirements for transparency, accountability, and ongoing support. Include community engagement requirements in your implementation process.

• Start small and measure carefully. Pilot AI tools in low-risk environments where you can evaluate performance without affecting critical operations. Use these pilots to develop institutional knowledge and refine your policies.

The goal isn't to avoid AI—these tools offer substantial benefits for public safety operations. Instead, you can ensure that AI adoption enhances rather than undermines your core responsibility of protecting and serving your community effectively and equitably.

You can watch the workshop here. 

The full series on AI and Law Enforcement is here.

The AI and Law Enforcement Workshop Series is cohosted by InnovateUS in partnership with the State of New Jersey, the AI Center of Excellence for Justice, Public Safety, and Security, and the Rutgers Miller Center on Policing and Community Resilience.