New Jersey’s former attorney general on Ring cameras and facial recognition

Episode Summary

Executive Summary: Anne Milgram argues that data can make policing and criminal justice more effective, but only if it is collected, audited, and used carefully to avoid reinforcing existing bias. She contrasts data-driven reductions in crime with the risks of surveillance, facial recognition, and private tech companies shaping law enforcement without clear rules, while stressing that public safety should mean preventing future crime, not simply increasing arrests and incarceration.

Main Topics: Data-driven policing and crime reduction (Priority: 5/5): Milgram explains how data and analytics transformed policing in Camden, New Jersey by revealing hotspots, deployment patterns, and crime networks, enabling a major reduction in violence. Bias in criminal justice data and decision-making (Priority: 5/5): She warns that criminal justice data often reflects historical racial and socioeconomic bias, and that the system must first measure its own practices before relying on the outputs. Reframing effectiveness and public safety (Priority: 5/5): Milgram argues effectiveness should be defined by preventing future crime, not maximizing arrests or incarceration, and notes that detention often worsens outcomes for low- and moderate-risk people. Surveillance, Ring cameras, and private-public data sharing (Priority: 4/5): The discussion examines how private surveillance tools create new flows of information to police, raising concerns about marketing, consent, and the normalization of pervasive monitoring. Facial recognition and regulatory guardrails (Priority: 4/5): She says facial recognition may have legitimate uses but needs clear limits, democratic oversight, and ongoing regulation that can adapt as technology changes. Institutional trust, federal dysfunction, and state action (Priority: 3/5): Milgram closes by stressing the importance of DOJ, courts, and states as checks on federal power, and says states are increasingly central to privacy and justice reform.

Key Arguments: Criminal justice systems should start by collecting basic operational data so agencies can understand who is entering the system, for what reasons, and how time and resources are actually spent. Data should be used to improve judgment, not replace it; the best model combines analytics with human experience and accountability. Most people in the criminal justice system are low- or moderate-risk, and incarceration often increases their likelihood of future crime rather than reducing it. Racial bias and mental-health-related disparities already exist in the system, so the first task is to measure those disparities honestly before deploying more advanced tools. Not all data should be used for every purpose; high-bias inputs such as misdemeanors or drug arrests may need to be excluded from predictive tools. Private surveillance products can be useful to law enforcement, but they should be governed by explicit rules about what is shared, when it is shared, and for what purpose. Facial recognition may support specific investigations, but widespread or open-ended surveillance would be a major civil-liberties threat and should not be deployed without clear legal guardrails. Police departments often buy technology they do not fully understand; departments should build internal data capacity so they become smarter buyers and users of technology. Public safety reform should focus on preventing crime upstream, including screening for mental health, substance abuse, and homelessness where those conditions repeatedly cycle people back into police contact. Federal institutions matter because they constrain executive power, but states will continue to drive much of the practical action on privacy, criminal justice, and tech regulation.

Data Points: Camden violent crime reduction: over 40% - Milgram says violent crime fell by more than 40% in one year after the police department was redeployed using data. Camden homicide reduction: 46% - She cites a 46% drop in homicides in Camden after the data-driven overhaul. Camden population: less than 80,000 - Used to emphasize how unusually high 911 call volume was for a relatively small city. 911 calls in Camden: over 11,000 to 12,000 - Milgram says Camden had more than 11,000–12,000 911 calls when she arrived. Date of AG swearing-in / iPhone launch: June 29, 2007 - She notes she became New Jersey AG the same day the iPhone launched. High-risk share of criminal justice population: about 10% - She estimates high-risk individuals make up only a small fraction of people in the system. Low-risk share of criminal justice population: about 50% - Her rough breakdown suggests roughly half of the population is low-risk. Moderate-risk share of criminal justice population: about 40% - Her rough breakdown suggests around 40% are moderate-risk. Incarceration impact on low/moderate risk: more crime - She says detaining low- and moderate-risk people leads to more pretrial and long-term crime. States in the U.S.: 50 - Used in the discussion of why state-by-state tech regulation can create a fragmented legal landscape. Approximate number of police departments in the U.S.: 20,000 - Milgram cites the large number of departments to explain uneven data capabilities nationwide. Police departments with more sophisticated capacity: 10 massive departments, maybe 20 - She says the biggest departments are generally better at using data than the thousands of small ones. People of color detained longer: 8.5 days longer on average - From a data pull in one city, matched on relevant factors. Mental illness disparity: significantly longer in jail; twice as bad outcomes - She says people with mental illness spent significantly longer in jail, even after matching on age, gender, and crime. Police department size: most are under 10 people - Used to explain why many departments cannot implement sophisticated processes on their own. Technology budget example: $3 million to $4 million bids - Outside vendors proposed costly Camden tech systems. Internal Camden build cost: under $1 million - Her team built a custom system internally for far less than vendor bids. Ring-style home camera use: external and internal cameras; baby monitor mentioned - Discussed as an example of voluntary private surveillance in everyday life.

Pivotal Quotes: "public safety is preventing future crime" — Anne Milgram: She defines the proper goal of the criminal justice system. "the bias is baked in right now" — Anne Milgram: She warns that existing data reflects long-standing racial and institutional bias. "we should make sure it doesn't happen" — Anne Milgram: She says surveillance overreach and misuse of data must be proactively prevented through clear rules and regulation.

Implications: Listeners should expect data and AI to expand in policing, but the real debate is about rules, auditing, and purpose. Without guardrails, surveillance can entrench bias; with them, data may help reduce crime and incarceration at the same time.