Machine guarding is the physical protection, barriers, interlocked covers, devices, that keeps body parts out of a machine's dangerous zones. OSHA's general-industry rule, 29 CFR 1910.212 requires one or more guarding methods on any machine whose point of operation, nip points, rotating parts, or flying chips could injure someone.
It is one of the shortest standards OSHA enforces and one of the most cited, because the failure mode is so ordinary: a guard removed for a jam and never rebolted, a new fixture that opened a gap, a machine that arrived from the used market with nothing on it at all. This guide covers what the standard actually says, where machine injuries happen, the four guard types and when each fits, and how guarding sits inside the hierarchy of controls.
What does OSHA 1910.212 actually require?
The standard requires guarding wherever a machine can hurt someone, and it names the hazards directly. Under 1910.212(a)(1), one or more methods of machine guarding must protect the operator and other employees in the machine area from hazards such as those created by the point of operation, ingoing nip points, rotating parts, and flying chips and sparks, with barrier guards, two-hand tripping devices, and electronic safety devices given as examples.
Four more requirements matter on the floor. Guards must be affixed to the machine itself where possible, and secured elsewhere if not, and a guard must not create a hazard of its own (a)(2). The point of operation, the area where work is actually performed on the material, must be guarded on any machine that exposes an employee to injury there, per (a)(3)(ii), and the guarding must prevent the operator from having any part of the body in the danger zone during the operating cycle. Fan blades less than seven feet above the floor or working level must be guarded, with openings no larger than one-half inch, per (a)(5). And machines designed for a fixed location must be securely anchored so they cannot walk or tip, per 1910.212(b).
Specific machine families, mechanical power presses, woodworking machinery, power-transmission apparatus, have their own standards (1910.217, 1910.213, 1910.219). But 1910.212 is the catch-all: if a machine can injure someone and no specific standard covers it, this one does.
Where do machine injuries actually happen?
Machine injuries cluster in four zones, and every guarding review starts by finding them on your machine.
- Point of operation where the cutting, forming, punching, or shearing happens. Highest severity: this is where amputations occur.
- Power transmission belts, pulleys, chains, sprockets, couplings, flywheels, and gears between motor and work.
- Ingoing nip points anywhere two parts move together (rollers, belt-and-pulley, rack-and-gear) and can draw in a hand, glove, or sleeve.
- Ejected material flying chips, sparks, and broken tooling leaving the machine at speed.
How bad is the amputation problem?
Bad enough that OSHA runs a standing National Emphasis Program on it. The numbers behind the enforcement:
- BLS data cited in OSHA's National Emphasis Program on Amputations in Manufacturing Industries counted roughly 6,200 work-related amputations with days away from work in 2018 and 58 percent involved machinery.
- Machine guarding (1910.212) was No. 10 on OSHA's most-cited list for FY 2024, with 1,541 citations (OSHA Top 10 Most Frequently Cited Standards).
- For context, private industry overall recorded 2.3 total recordable cases per 100 full-time workers in 2024 (BLS), but amputations carry some of the longest recovery times of any injury class.
The NEP means inspectors specifically target plants with machinery known to cause amputations, presses, saws, shears, slicers, and plants previously cited under the guarding and lockout/tagout standards.
What are the four types of machine guards?
OSHA's guidance groups guards into four types. The right one depends on how often the operator needs access and how much the stock size varies.
Guards are not the only method the standard accepts. Two-hand tripping devices keep both hands on controls and out of the die. Presence-sensing devices (light curtains) stop the cycle when something breaks the plane. Pullbacks and restraints physically tether hands out of the point of operation. These devices count as guarding methods under 1910.212(a)(1), but a device that depends on maintenance and calibration needs verification on a schedule, not faith.
How do you review a machine for guarding? A six-step check
Run this on every machine, new, moved, or modified, and on a recurring inspection cycle.
- Map the hazard zones. Find the point of operation, every power-transmission component, every nip point, and every direction the machine can throw material. A job safety analysis for the operator's tasks will surface zones a walk-around misses.
- Test reach, not appearance. A guard passes if nobody can reach the danger zone over, under, around, or through it during the cycle, not because a barrier merely exists.
- Check the guard itself. Affixed to the machine, no sharp edges or new pinch points, doesn't force the operator into an awkward posture that invites bypassing it.
- Verify interlocks function. Open each interlocked guard with the machine running its cycle in a controlled test: it must stop. Log the test.
- Confirm the boundary with maintenance work. Guarding protects normal operation; the moment someone defeats a guard to clear a jam or service the machine, you are in lockout/tagout territory, and the energy must be isolated.
- Train and re-inspect. Operators need to know what each guard does and how to report a damaged or missing one. Missing-guard checks belong on shift-start checklists and in your safety audits.
Where does guarding sit in the hierarchy of controls?
Guarding is an engineering control, below eliminating the hazard, above rules and PPE. That ordering is the answer to a common floor argument: gloves and training do not substitute for a guard, because they depend on a human doing the right thing every cycle, and the machine only needs one exception.
Why do guards fail in practice?
Almost never because the guard was engineered wrong. They fail administratively: removed during maintenance and not replaced, interlocks taped or jumpered to speed up jam clearing, guards left off equipment bought used, or nobody assigned to check. Which means the fix is mostly a visibility problem. Guards need to be on a named inspection checklist, missing-guard reports need a frictionless path from operator to fix, and defeated interlocks need to surface the day they happen, not at the annual audit. Plants that digitize those checks and reports, the same connected worker move that replaces paper forms with tablets at the station, get a live record of guard status per machine instead of a binder that's current once a year. A five-minute toolbox talk on why an interlock exists, backed by data on how often it's found bypassed, beats a policy memo every time. That's the pattern Harmony builds on: capture the check where the work happens, and make the exception visible to the person who can fix it (see how it works).