AI in policing risks amplifying discrimination, researchers

Police forces deploying AI risk encoding and amplifying existing discrimination against minorities, according to researchers raising concerns about algorithmic bias in law enforcement.

The core problem is structural: AI systems trained on historical police records – which themselves reflect decades of discriminatory enforcement patterns – inherit and sometimes magnify those biases. An algorithm trained on arrest data that overrepresents minority communities will flag similar populations at higher rates, creating a feedback loop that justifies further disproportionate policing.

Without meaningful oversight, this becomes a governance failure. Police departments often lack transparent audit trails for algorithmic decisions, making it impossible to identify when or how bias enters the system. The researchers emphasise that the issue isn't that AI is inherently discriminatory – it's that organisations deploying these tools without adequate due diligence, bias testing, or external accountability are gambling with public trust and civil rights.

This matters beyond policing. The pattern – deploying decision-making technology trained on biased historical data, then claiming objectivity – appears across hiring algorithms, credit scoring, and benefit eligibility systems. Each one perpetuates systemic inequality under the guise of neutrality.

The gap isn't technical; it's governance. Organisations need mandatory bias audits before deployment, ongoing monitoring, and independent verification – not as optional extras, but as non-negotiable requirements. Where are police forces on this? And what accountability mechanisms exist when harm occurs?

Police using AI risks making discrimination worse, researchers warn

Report This Content