Taser maker Axon is mandating AI use to speed up work as its surveillance tools face scrutiny

When Rick Smith founded Taser International in an Arizona garage in 1993, his original vision involved taking a device that was originally created by a NASA scientist and redesigning it in a way that would make it palatable to law enforcement agencies.

More than three decades later, Smith’s company—now called Axon Enterprise—has sold more than one million Taser devices and diversified its portfolio to include body cameras, drones, and license plate reader cameras to law enforcement, emergency responders, and private-sector security customers. 

More recently, Axon has also focused on selling adjacent software and artificial intelligence tools, helping the company report ten consecutive quarters of revenue growth above 30%. But its push into AI coincides with escalating criticism about how officers are using these technologies for surveillance. 

Police officers have been either fired or charged with using Flock Safety’s AI-enabled cameras to stalk former romantic partners and other personal connections. AI systems erroneously led to an arrest of a man accused of trying to lure a child from a McDonald’s in 2023, and more recently, led to a Florida woman spending two weeks in jail after being wrongly accused in a deadly hit-and-run incident.

Most of the public’s ire has centered on Flock. Axon has also faced criticism that its AI tools that help officers write police reports make errors, though the company said this occurred during an early access period when agencies could provide feedback before the product’s broader launch. And earlier this month, the Los Angeles City Council voted to delay a pilot program with Axon that would have attached AI-enabled license plate readers on streetlights.

In August, Smith acknowledged that there was a budding anti-surveillance movement that echoed the massive backlash that followed the release of Tasers. “Any new technology in public safety is going to go through an adjustment period, and I would say the critics aren’t all wrong,” Smith told investors during the company’s second-quarter earnings presentation.

Jeff Kunins, who holds the dual roles of chief product officer and chief technology officer at Axon, tells Fortune that the public discourse around surveillance and AI is a complex issue that requires a nuanced conversation.

“I am personally a life-long, die-hard progressive, and I’m here because of that, not in spite of it,” Kunins said. “We view our work as using technology to improve transparency and accountability on both sides…and our role is to build the tech that facilitates public safety.”

When Axon develops new features, Kunins says a third-party independent group, called the Ethics and Equity Advisory Council, makes recommendations during the product design and development process.

He says that when Axon developed Draft One, which utilizes generative AI to take body camera audio and turn it into a first draft of a police report, there were instances where the AI-generated transcript introduced bias based on a subject’s ethnicity. The large language models would generate text that said a white subject “left the scene,” while a Black person “fled the scene,” even when all other factors in the case were identical.

“Obviously, that’s a flagrant example of inappropriate bias coming out of the models solely because of the description of the skin color of the person,” says Kunins. Axon fine-tuned the models to ensure the outputs were more consistent and didn’t introduce harmful bias.

Other AI tools that the company has deployed include Axon Assistant, a tool that can translate more than 50 languages and dialects spoken within the body camera, allowing officers to communicate in the field without needing a human interpreter.

For nearly a decade, Axon has also used AI to help police departments redact data from camera footage, blurring out the faces of bystanders, license plates, and other sensitive details that aren’t involved in an active case. Previously, a law enforcement employee would need to manually go through every frame of a video to obscure those personal details before the footage could be shared with lawyers, the press, and the public.

Internally, Kunins has encouraged collaboration between the engineering team and the broader employee base through hackathon training sessions. The IT department is also tasked with customizing some AI features from software vendors, including Workday and Salesforce, to automate workflows.

There is an even more assertive effort to promote AI usage among Axon’s research and development team, which totals about 2,200 employees. Kunins says that by the end of 2026, Axon expects all work across product management, design, and software engineering to be compressed by 25% to 30%, requiring less labor and organizational capacity than would be needed if the same projects were developed without AI.

To get to that target, Kunins says he’s embraced a carrot and stick approach. Every team has a 10% mandatory target in productivity gains, meaning the baseline requires everyone will have fewer resources to advance new features and products. If teams can show they’ve accelerated those gains by more than 15%, Axon will award them stock grants, as long as the broader team achieves that 10% minimum threshold, according to Kunins.

“We’re expecting to yield about 25%-ish acceleration, not just of the coding part, but of the entire person effort going into these features,” says Kunins.

John Kell

This story was originally featured on Fortune.com