Axon’s Case for Safer Policing—and the Power Its Technology Creates

1. Guest Background

Rick Smith, the CEO of Axon, appears on this All-In Podcast interview as an executive whose experience spans electric weapons, police adoption, and the difficult construction of a software business inside a hardware company. Axon was formerly called Taser. The episode examines how that business developed and what its expanding capabilities mean for freedom and public safety. Smith supplies an operating perspective, but also speaks as the leader of a company selling the technologies under discussion.

He says he began the business in a Tucson garage in 1993, when he was 23. A conversation during graduate study in Europe prompted him to count people he knew who had been shot and killed, including two high-school football teammates. His founding question was whether a threatening person could be stopped without a bullet. Through patent filings, he found the original Taser inventor, a retired NASA scientist whom he persuaded to collaborate. Smith recalls a partnership between a 23-year-old and a 73-year-old, with an initial prototype assembled from hardware-store parts in 30 days.

His relevant experience includes acknowledging where that background stopped being sufficient. When the company entered body cameras and software, Smith says he struggled to build the right team. Board member Hadi Partovi effectively became a part-time CTO for about five years. That admission matters: the interview concerns both the invention of tools and the organizational capacity required to operate consequential infrastructure.

2. What the Episode Covers

The episode follows a change in what Axon sells: first an alternative means of stopping someone, then a record of police encounters, and ultimately a connected public-safety platform. Smith originally targeted consumers. Compressed air, replacing gunpowder, was intended to avoid the original device’s firearm classification and make purchasing easier. Yet consumers asked why police were not using it. He says that credibility problem helped drive the late-1999 pivot toward policing. His proposed return to consumers within five years remains a forecast, conditional on demonstrating performance comparable or superior to a gun in police use.

The pivot depended on financing as much as market insight. Smith recalls approximately $2 million in bank debt, with his father exposed to repayment. His father and investor Bruce Culver each supplied a final $500,000. According to Smith, that money carried the police product through design, tooling, manufacture, and launch. A former Marine then toured departments in a recreational vehicle, offering officers $100 if they could overcome the demonstration. Smith associates the tour with purchase orders and cash-flow break-even within about a year. He dates the IPO to 2001, at roughly a $12 million valuation.

Success did not eliminate institutional vulnerability. Smith says an SEC investigation beginning in late 2004 caused police customers to suspend buying, and that it ended with no findings. He then describes extensive litigation and twice arranging to take over insured liabilities so the company could contest claims rather than allow the insurer to settle. He also says actual wrongdoing can justify settlement. These are his accounts of investigations and litigation; the interview does not supply the underlying records needed to independently establish their scope or outcomes.

Body cameras changed the company’s problem from intervention to evidence. Smith traces the idea to a custody-death lawsuit in which he believed restrained courtroom descriptions failed to communicate what officers encountered. He wanted a visual record of what they saw and did. He identifies calls for cameras from Michael Brown’s family and a police union after Ferguson as a demand turning point. His argument connects public accountability with protection for officers facing false allegations.

The later discussion moves into drones and robotic intervention. Smith says Axon bought almost half of Skydio, preserving its independence while obtaining exclusive channels into selected markets. His expectation of millions of drones in the United States is a prediction, not an established deployment count. The closing discussion reveals the broader outlook behind these investments: Smith anticipates serious risks from weakening policing and possible unrest during technological transition, alongside substantial improvements over the next 10–20 years. His historical analogies and causal claims are not independently verified in the supplied interview.

3. Core Views: Reasoning, Examples, and Limits

Smith’s most concrete product argument starts with a failure. Early Tasers, he acknowledges, depended on pain, and determined, aggressive people could resist them; some of his own demonstrations failed. He says the team used animal testing to adjust nerve stimulation toward incapacitation. The proposed mechanism matters: inducing discomfort and preventing continued movement are different objectives. The touring $100 challenge then made the revised capability immediately legible to police buyers. But a demonstration establishes only what happened under its conditions. Pain, temporary incapacitation, successful resolution, and safety remain separate outcomes. Neither a persuasive sales encounter nor the company’s mission establishes performance across people and circumstances.

His account of body cameras similarly turns on a change in how users perceive the tool. Officers initially feared supervisory scrutiny. Smith recounts a union leader’s observation that officers resisted cameras until wearing them for 90 days, after which they would not patrol without one. His Phoenix example explains the claimed benefit: an officer told him an arrested woman threatened a rape allegation, then changed her behavior when informed the encounter was recorded. This illustrates how recording might protect an officer as well as expose misconduct. It does not establish the frequency of false complaints or the overall balance of benefits and harms. The episode supports a dual-accountability rationale more firmly than a general outcome claim.

The software transition adds another requirement: evidence must remain available and securely usable. Smith credits Partovi with rebuilding the operation after technical and staffing difficulties, and describes a data-center outage that nearly lost all customer data. Migration to AWS and later Azure addressed an infrastructure choice; the Axon name addressed the limits of an electric-weapon brand. Those are distinct changes. His claim that security depends on technology and people rather than a server’s physical proximity is a useful evaluative principle, but does not make every cloud implementation secure. Likewise, his million-officer upload estimate and comparison with YouTube lack independent verification and a common measurement basis. Scale is not a substitute for reliability evidence.

The surveillance debate exposes a difference in the questions being answered. Smith distinguishes legitimate investigations—such as finding stolen vehicles or examining plates near violent crimes—from private searches and universal tracking. He describes case-code validation, cooperation across systems, and another officer’s involvement. For Ring footage, he emphasizes homeowner-controlled sharing and restrictions on random police collection, using his difficulty retrieving footage after a car theft to explain the convenience benefit. These controls address specific access decisions. The host asks whether accumulated capability itself gives the state overwhelming power, even if individual searches have plausible purposes. The Minority Report analogy explores that boundary; it is not evidence that the systems described already conduct predictive arrests.

Drones bring the same disagreement into physical intervention. Smith claims rapid arrival from autonomous docks, sometimes before a 911 call is fully entered, and cites distraction and surrender incidents. Public mission profiles and incident-triggered flights are his proposed constraints. These examples explain potential uses without establishing aggregate savings or reduced harm. His defense of continuing Taser-drone research is a mechanism argument: remove the operator from immediate personal danger, and nonlethal intervention may become more feasible than sending armed people into a confrontation. That possibility deserves comparison with present police practice, while the host’s concern requires a second comparison: how much additional intervention power would deployment create?

Smith proposes keeping force decisions human, potentially requiring real-time administrator approval, and subjecting robotic use to independent civilian review. His mistaken-entry scenario contrasts a remote team with armed officers making rapid decisions while fearing for their lives. It provides a reason to investigate the design, not proof that multiparty authorization will reduce error. Deployment thresholds and responsibility still need definition. Smith also says his reconstituted ethics council includes more affected-community voices and argues criticism should enter before designs harden. His explanation that departing advisers faced political pressure remains his interpretation, not their testimony. Similarly, his claims about poorer communities welcoming police technology and apprehension probabilities outweighing sentence length are assertions for evaluation; the episode supplies no supporting dataset.

4. Learning and Application

For product teams and purchasers, the first practical use is to separate the outcomes hidden inside the word “effective.” Smith’s account of pain-based failure suggests asking whether a device produces discomfort, incapacitates a person, resolves the encounter, and causes harm. Those questions require different evidence. A demonstration can help a professional buyer understand a capability; deployment decisions also need defined failure conditions and performance under relevant circumstances. The consumer demand for police adoption shows how institutional credibility can influence market sequence, but institutional purchasing should not be treated as a replacement for verification.

Business analysis should also test whether procurement survives a loss of confidence. Smith’s final financing round supported a launch, while the later investigation, in his account, interrupted police purchases. A comparable company could use those experiences to model purchasing delays, maintenance obligations, and cash available during disputes. His insurance strategy should remain a case-specific account rather than a rule that litigation must always be fought. He explicitly preserves settlement when the company has done something wrong; the interview does not provide enough case detail to recommend reproducing his decisions.

For hardware companies taking responsibility for evidence, software staffing and recovery capability belong in the delivery plan. Partovi’s intervention and the near-loss of customer data suggest concrete review questions: who owns recovery, what has been tested, and who controls access and sharing? Cloud services may reduce the burden of operating a data center, but the assessment still needs to examine personnel, configuration, and responsibility. Neither “local” nor “cloud” answers those questions by itself. A broader brand can explain a broader business, while operational competence must support the promise.

Public agencies can translate Smith’s safeguards into an inspectable chain of decisions. Before a plate search, establish its case basis. For household footage, specify the homeowner’s sharing control. For drones, define the triggering request, destination, and mission information available to the public. Before robotic force, identify the human decision maker and whether another authorization is mandatory. After use, specify the independent civilian review. Each stage should make clear who can refuse, how exceptions are detected, and what happens when a requirement is unmet. These are proposed ways to evaluate his framework, not claims that every described control is already implemented everywhere.

Finally, a pilot should evaluate two baselines together: the risks of the current response and the additional tracking or intervention enabled by the new system. Incident-triggered flight does not settle retention or later use of collected information. Multiple approvers may create useful friction while still requiring clear responsibility. Early ethics consultation should therefore document whose concerns changed which design choices. Reporting should also distinguish anecdotes, scale estimates, and forecasts: a surrender incident is not an aggregate outcome, upload volume is not governance maturity, and a projected drone market cannot justify present permissions. Those boundaries preserve the episode’s useful mechanisms without promoting its promises into established results.

Source

• Original episode: Freedom vs. Safety: Axon CEO Rick Smith on the Future of Policing