Imagine a police officer using an AI assistant to analyze crime scene photos or draft reports – sounds like science fiction becoming reality. Artificial intelligence startup Perplexity has launched a new program offering its Enterprise Pro tier to public safety organizations, including police departments, for free for one year. This move marks a significant expansion of AI into law enforcement, but it raises critical questions about accuracy, accountability, and the real-world consequences of algorithmic errors in high-stakes environments.
The Promise and Peril of Police AI
Perplexity’s “Public Safety Organizations” program, unveiled in January, allows police to use AI for tasks like analyzing crime scene photos, processing body camera transcripts, and generating structured reports from investigators’ notes. The company positions this as helping officers make more informed decisions in real time and automating routine administrative work. But what happens when AI makes subtle errors in police reports that could lead to wrongful convictions?
“What can be pernicious about these kinds of use cases is they can be presented as administrative or menial,” says Katie Kinsey, chief of staff and AI policy counsel at the Policing Project. “There’s a lot of important decision-making, leading to charges and indictments, that emanates from the kinds of use cases they’re talking about here.”
The Hallucination Problem in High-Stakes Contexts
AI chatbots are notoriously prone to “hallucinations” – generating plausible-sounding but false information. While an AI fabricating a story about an officer shape-shifting into a frog might be easily dismissed, more dangerous scenarios involve subtle alterations of truth that are difficult to detect. Recent testing by ZDNET revealed that when asked about recent news stories, Perplexity and other leading chatbots frequently generated responses with “at least one significant issue” related to accuracy or sourcing.
Andrew Ferguson, a professor at George Washington University Law School, emphasizes the gravity of the situation: “When you are playing with liberty and constitutional rights, you need to make sure that safeguards are in place for accuracy � without laws or rules to protect against mistakes, it is incumbent on the police to make sure they use the technology wisely.”
The Regulatory Landscape: State vs Federal Approaches
As AI adoption accelerates in law enforcement, regulatory frameworks struggle to keep pace. California’s SB-53 and New York’s RAISE Act, both effective in early 2026, represent state-level attempts to regulate AI safety. These laws require AI model developers to publicize risk mitigation plans and report safety incidents, with fines up to $1 million in California and $3 million in New York for non-compliance.
However, these state regulations face federal pushback. The Trump administration has renewed attacks on state AI legislation through an executive order and an AI Litigation Task Force, arguing that state regulations create a patchwork that stifles innovation and could cede ground to China. Gideon Futerman, special projects associate at the Center for AI Safety, notes: “SB-53’s level of regulation is nothing compared to the dangers, but it’s a worthy first step on transparency and the first enforcement around catastrophic risk in the US.”
The Broader Business Context: AI Deployment Outpaces Safety
Perplexity’s move into law enforcement reflects a broader trend where businesses are deploying AI faster than safety protocols can keep up. A Deloitte report reveals that while 23% of companies currently use AI agents moderately – projected to jump to 74% in two years – only 21% have robust safety mechanisms. This rapid adoption trajectory creates significant limitations as AI scales from pilots to production deployments.
The Deloitte report authors warn: “Given the technology’s rapid adoption trajectory, this could be a significant limitation. As agentic AI scales from pilots to production deployments, establishing robust governance should be essential to capturing value while managing risk.” They recommend implementing oversight procedures, clear boundaries for agent autonomy, real-time monitoring, and audit trails.
Who Bears Responsibility?
The central question remains: Who should ultimately be responsible for ensuring AI is used responsibly within law enforcement? Kinsey believes the onus should fall on policymakers: “The problem is there’s no hard law that’s setting out what these requirements should be.” Meanwhile, Perplexity claims it’s well-positioned to equip public safety personnel with AI tools since it’s made accuracy a key part of its product and business model.
This tension between technological capability and regulatory oversight mirrors broader challenges in AI governance. As businesses across sectors rush to implement AI solutions, the gap between deployment speed and safety protocols widens, creating potential risks that extend far beyond law enforcement.
The Future of AI in Public Safety
While Perplexity says this is the first program of its kind, it almost certainly won’t be the last. AI developers face huge pressure to expand their user bases, and police departments have a long history of being early adopters of new technologies. From “predictive policing” algorithms in the early 2000s to current uses of facial recognition and lie detection, law enforcement continues to embrace technological solutions.
“Law enforcement is a good client to have, because they’re not going anywhere,” notes Kinsey. “We see that relationship between private industry and law enforcement all the time.” The intense competition of the AI race could result in other companies following Perplexity’s lead by launching initiatives aimed at police officers and other public safety officials.
As AI becomes increasingly embedded in critical decision-making processes, the need for robust governance frameworks, clear accountability structures, and ongoing evaluation of accuracy becomes paramount. The conversation about AI in law enforcement isn’t just about technology – it’s about how we ensure that technological advancement doesn’t come at the cost of justice, accuracy, and public trust.