SAFE
Patrol work is responding to calls in unpredictable physical settings, physically detaining people, controlling traffic scenes, and making use-of-force and arrest decisions in seconds — none of which current AI or robotics can do. The automatable slice is real but administrative: incident report drafting, citation paperwork, records queries, and license-plate and camera review. Only a sworn, state-certified officer can make a custodial arrest or execute a warrant, and that officer is personally named in the litigation that follows.
Roughly flat across the period, with year-to-year wobble.
Median pay $63,150 → $76,210 -3.5% in real terms
This line is counted by the Bureau of Labor Statistics — the one figure on this page that isn't a judgement of ours. Headcount moves on demand, offshoring, demographics and the business cycle, and automation is one term among several, often not the loudest.
So a falling line is not evidence that AI did it, and a rising one is not evidence that it won't. Both happen in this register: some occupations resist automation and shrink anyway, others are highly automatable and keep growing. The marked year is 2020.
BLS projection, 2024–2034
+3.1% 698,800 → 720,800 on the projections basis
Hard to automate, and growing
The work resists current AI and the BLS projects +3.1% more of these jobs by 2034. Note that safe does not mean well paid — several of the fastest-growing resistant occupations are among the lowest paid on the register.
Different clocks. The score is what current AI could do to this work today. The projection is how many of these jobs will exist in 2034. Everything between the two — how fast employers actually adopt, whether demand grows in the meantime — is why they can point opposite ways without either being wrong.
~53,700 openings a year on average, including replacing people who leave.
CopDeputyMarshalOfficerTrooperConstablePatrolmanPolicemanAnimal CopPark PolicePolicewomanRoad DeputyBorder GuardCity MarshalDesk OfficerDuty OfficerPatrol DeputyPatrol DriverPeace OfficerPolice WorkerSpecial AgentState TrooperUnder SheriffCity Constable
Holding it up: embodiment . Weakest point: trust premium .
Tasks largely resist digitisation A patrol shift is domestic disturbances, welfare checks, foot pursuits, roadside sobriety testing and crash scene control — tasks that require entering a stranger's living room and reading a situation in real time, which is why the score sits at 16 rather than 20: report narratives, e-citations, NCIC/plate queries and evidence logging are already being drafted and auto-populated by software.
Hands-on in uncontrolled environments Officers work outdoors in weather, at night, on highways with live traffic, and go hands-on to handcuff resisting subjects, drag victims from vehicles and clear structures — there is no controlled environment anywhere in the job, which is what a 20 means.
Licensed human required and personally liable State POST certification is a legal prerequisite to make a custodial arrest or swear a warrant affidavit, and officers are personally named as defendants in §1983 excessive-force suits with qualified immunity decided case by case; it stops short of 20 only because qualified immunity and municipal indemnification absorb most of the financial hit.
Exists to be accountable for ambiguous calls Whether a Terry stop is supported by reasonable suspicion, whether force is objectively reasonable under Graham, whether to arrest or divert a person in mental health crisis — these are seconds-long, irreversible calls made on incomplete facts and reviewed later by prosecutors, courts and internal affairs, with department policy narrowing but never resolving the ambiguity.
The verdict above describes this occupation as a whole. Almost nobody does the typical version of a job — tick what's actually in your week and see how your own mix sits.
Your task mix speaks to task resistance (16/20 here) — how much of the day's work current AI already does. That is the dimension the boxes above are about.
It cannot move the other three. Liability shield (16/20) is whether the law requires a licensed human to sign. Trust premium (13/20) is whether buyers specifically pay for a person. Judgment and accountability (17/20) is whether the role exists to own consequential calls. Those are facts about the occupation's standing, not about which tasks are in your week — a paralegal who does only trial exhibits still holds no licence. Together they are 46 of this occupation's 82 points (56%).
Embodiment (20/20) is also a property of the work rather than the worker, but we don't tag individual tasks as physical or not, so the picker can't tell you anything about it. That's a limit of this tool, not a claim.
Did we get the list right? Tell us what's missing — the tasks are written from the outside, and you're reading this from the inside.
The moves above are yours to make. This is the other half: what would have to change in the world for the occupation itself to score higher. None of it is in any one person's gift, but it is where the floor actually comes from. Scores here are not a one-way ratchet. Only two of the five dimensions — task resistance and embodiment — track what machines can do. The other three track law, what buyers will pay for, and who is answerable, and those move in both directions, often in response to the same pressure AI creates. If every lever below landed, this occupation would score around 91/100, still SAFE.
State POST (Peace Officer Standards and Training) boards or legislatures explicitly barring AI-generated probable-cause narratives from being submitted without a sworn officer's personal attestation — already in motion: Utah SB 180 (2025) requires disclosure when reports are AI-drafted, and King County WA prosecutors (2024) refuse Axon Draft One AI reports outright. If attestation-plus-personal-liability language spreads to statute, the sworn signature becomes the only lawful route to charging.
Post-Fourth-Amendment litigation over algorithmic tips (facial recognition, ShotSpotter, ALPR) settling into a rule that an officer must independently corroborate machine output before seizure — several circuits and city ordinances (e.g. Detroit's 2024 FR settlement with the ACLU) already require independent corroboration. That formally assigns the consequential call to the human.
Task-mix shift: if report drafting, citation entry, and camera/ALPR review are absorbed by tools, the residual job is almost entirely street contact, de-escalation, and force decisions — a genuinely two-tier occupation where the automatable tier is the smaller and lower-judgment one. Raises resistance without any new law, though it also reduces headcount need per call volume.
Union contract or consent-decree language requiring in-person human response for specified call types (domestic violence, welfare checks) rather than remote/automated triage — FOP and IUPA locals have bargained against civilianization and drone-as-first-responder programs in Chula Vista CA and NYPD. Community-oversight boards demanding human presence are the visible mechanism.
The limit. Already 82/100; headroom is ~8 points at most and embodiment is maxed. The realistic risk direction is not displacement but headcount compression per agency as drone-as-first-responder and AI dispatch triage reduce officers needed per call — a scope change the register's dimensions do not capture.
| New York-Newark-Jersey City, NY-NJ | 54,270 | $105,540 +38% |
| Los Angeles-Long Beach-Anaheim, CA | 24,720 | $121,980 +60% |
| Chicago-Naperville-Elgin, IL-IN | 22,680 | $107,810 +41% |
| Dallas-Fort Worth-Arlington, TX | 16,430 | $92,910 +22% |
| Houston-Pasadena-The Woodlands, TX | 16,250 | $74,570 -2% |
| Miami-Fort Lauderdale-West Palm Beach, FL | 14,710 | $79,840 +5% |
| Washington-Arlington-Alexandria, DC-VA-MD-WV | 14,630 | $84,000 +10% |
| Philadelphia-Camden-Wilmington, PA-NJ-DE-MD | 13,160 | $89,580 +18% |
| San Jose-Sunnyvale-Santa Clara, CA | 2,590 | $136,140 +79% |
| San Francisco-Oakland-Fremont, CA | 7,990 | $132,000 +73% |
| Carson City, NV | 110 | $131,210 +72% |
San Francisco Police Department
Hoodline reported that San Francisco police used a drone operation to identify and apprehend burglary suspects posing as a construction crew in the Northern District.
Has AI actually changed your work? One tap, anonymous, and the running tally is public. Nothing else is asked of you.
Rather than check back: get the digest and we'll tell you what changed — or watch a single occupation from its own page.