How the Federal Government Can Rein In A.I. in Law Enforcement

Harms of Law Enforcement's Use of A.I.

Law enforcement's use of emerging A.I. technologies has led to false arrests and police seizures, causing harm and consequences for innocent individuals. Facial recognition technology has misidentified several people, resulting in arrests. These errors often happen because of a lack of transparency, community agreement, and proper training in law enforcement agencies.

A 2018 study found that commercial facial-analysis programs showed biases in skin-type and gender, with darker skin leading to more errors. The use of such technologies raises questions of fairness and bias. Law enforcement must consider the consequences and ensure that the data sets driving these technologies don't contain errors or racial bias.

The proposed guidance by the Office of Management and Budget covers various law enforcement technologies, including facial recognition, license-plate readers, predictive policing tools, and social media monitoring. It emphasizes transparency, assessment of costs and benefits, and reduction of risks to individuals, especially in marginalized communities. However, stronger measures are needed to ensure compliance and prevent discriminatory use of these technologies.

Loopholes in the Proposed Guidance

While the proposed guidance is commendable, it leaves room for loopholes that could undermine its intended purpose. The provision allowing senior officials to seek waivers based on claims of hindrance to law enforcement operations is concerning. Clear proof should be required for any waiver, and there should be strict review to prevent misuse of this provision.

The vague exception for “national security” is another issue, as it can be conveniently invoked to evade legal protections and oversight. A sharper definition is necessary to ensure that the exemption is only used when valid cause exists. Additionally, the proposed guidance only applies to federal government agencies, leaving out state and local agencies that also use emerging technologies. Federal funding should be used as an incentive for compliance at all levels of law enforcement.

To truly address the potential harms of A.I. in law enforcement, stronger standards and measures must be put in place, with careful oversight and enforcement. The progress envisioned in the proposed guidance should not be undermined by backdoor exceptions and lack of compliance at different levels of government.

Conclusion

The federal Office of Management and Budget's proposed guidance on the use of A.I. in law enforcement is a step in the right direction, but it falls short in ensuring accountability and preventing discrimination. Stronger measures should be implemented, including stricter review of waivers, a clearer definition of national security exceptions, and extension of the guidance to state and local agencies. By setting a higher standard at the federal level, the government can lead the way in responsible and ethical use of A.I. technologies.

Joy Buolamwini and Barry Friedman, both experts in algorithmic justice and law, emphasize the importance of preventing the misuse of A.I. in law enforcement. They advocate for stricter measures and highlight the potential consequences for marginalized communities if these technologies continue to be used without proper safeguards. It is crucial for the Office of Management and Budget to prioritize accountability and compliance in its finalization of the guidance.