Quick Hits
- AI tools can be extremely helpful in supporting the need for constant safety oversight in manufacturing environments
- However, employers retain the responsibility to comply with OSHA regulations, even if the employer implemented an AI process designed to protect worker safety.
AI has many positive use scenarios in a manufacturing setting. A camera notices that a worker walked into a press area without safety glasses. A sensor on a conveyor motor picks up a vibration pattern that has previously indicated an impending failure. A forklift slows down as a pedestrian comes around a blind corner. In this AI age, manufacturing employers implement AI resources not only to increase efficiency, but to better promote workplace safety—striving for fewer injuries, fewer disruptions, and a stronger safety culture. But one principle applies to AI in any setting: AI works best as a tool that supports human judgment, not one that replaces it. Without real human oversight, AI safety systems can leave gaps that could lead to worker injuries and OSH Act compliance shortfalls.
How AI Is Supporting Worker Safety
Used well, AI can extend what a safety team is able to see, predict, and respond to. Applications of AI are growing each day, but common uses in manufacturing can include:
- Personal protective equipment (PPE) compliance: Cameras paired with AI models can flag missing hard hats, eye and face protection, or proper gloves.
- Safety alarms: Systems can watch the restricted areas around hazardous machinery or blind spots and send warnings when employees enter an area where collision could occur.
- Predictive maintenance: AI can analyze sensor data and spot signs of equipment wear or failure before a breakdown, reducing the need for emergency maintenance.
- Incident analytics: AI can sort through large volumes of incident and near-miss reports to find trends that would be hard to see manually.
- Collaborative robots: “Co-bots” can take over repetitive or hazardous tasks and reduce workers’ exposure to strain and repeat use injuries.
Each of these tools can make a safety program stronger, but relying too heavily on any of them can create inadvertent gaps in even the most well-intentioned safety program. These gaps create openings for unintended compliance shortfalls and—potentially—employee injuries. Employers may want to continue intentionally monitoring their safety programs—even those supported by AI tools—to identify these gaps and create solutions to fill them.
Why Human Oversight Still Matters
However, AI does not replace competent or qualified persons, hazard assessments, or enforcement of safety rule violations. If employees assume “the system will catch it,” basic safety practices can slip. And it is the employer that remains responsible: the duty to comply with the OSH Act’s general duty clause (Section 5(a)(1)) and applicable Occupational Safety and Health Administration (OSHA) regulations cannot be handed off to a vendor or an algorithm.
Mapping AI to Commonly Cited OSHA General Industry Standards
OSHA has not issued regulations specific to AI use in safety programs. However, the following OSHA general industry standards come up often in manufacturing and are among OSHA’s Top Ten Most Frequently Cited Standards for the most recently reported year. AI can assist with each, but people must still oversee these processes to ensure compliance with OSHA requirements.
| OSHA Standard | How AI Can Help | Where Human Oversight Is Required |
| Lockout/Tagout (LOTO) 29 CFR 1910.147 | Monitoring for energized equipment or missing locks | Energy control procedures, employee training and communication, and periodic inspections must be performed by people. AI-driven automated machinery must itself be covered by the LOTO program. |
| Machine Guarding 29 CFR 1910.212 | Vision systems and light curtains that detect intrusion into danger zones | AI tools can supplement required physical guarding but do not replace it. People must verify that guards are in place and that employees are trained on proper usage. |
| Powered Industrial Trucks 29 CFR 1910.178 | Proximity sensors and collision warnings | Operator training, evaluation, and certification remain required. |
| Hazard Communication 29 CFR 1910.1200 | Managing Safety Data Sheets (SDS) and labels; quick chemical information lookup | People must verify the accuracy and deliver required training. |
| Respiratory Protection 29 CFR 1910.134 | AI sensors can detect airborne hazards | Required written program, medical evaluations, and fit testing must be done by humans. |
The pattern is the same across the board: AI can improve detection and efficiency, but human involvement is necessary to verify the accuracy of written programs, conduct training, carry out inspections, and implement enforcement programs. After all, it is still the employer that will be liable for an OSH Act violation, even if the employer implemented an AI process designed to protect against those violations.
Top Five Tips for Effective Use of AI in Manufacturing Safety Systems
Employers thinking about adopting AI safety tools, or already using them, may want to consider the following tips:
- Robust safety teams. Assessing the current human safety team before implementing AI tools for any safety tasks is a critical early step. Employers may want to identify whether the team is comprised of the right people, with the right experience and knowledge of the workplace, to identify potential hazards and whether AI tools can assist in reducing those risks.
- Intentionality. Consider starting with a traditional hazard identification and assessment. This step includes intentionally identifying the hazards that may exist before implementing an AI process, as well as the hazards that could be caused by the AI tool itself, and implementing only AI processes that are necessary or are tied to a specific safety-related goal.
- Integration, not replacement. Because human oversight is crucial in all safety programs, it is important to note that AI safety tools are not a substitute for an employer’s existing safety and health programs. Human involvement could take many forms, including:
- Defining alert ownership. Factors include deciding who reviews alerts, how quickly they must respond, and what they are expected to do.
- Validating and auditing performance. Standard elements in this step include testing systems under real site conditions, and auditing them again after changes to equipment, layout, or processes.
- Keeping a human in the loop for stop-work decisions. Employers may want to remind their employees that people, not AI, should make the final call on whether to stop or restart work.
- Training workers and supervisors. Employers may want to ensure that workers and supervisors understand what the tools do, where they fall short, and why ordinary safety practices still apply.
- Documenting oversight activities. Keeping records of validation, audits, alert responses, and corrective actions demonstrates that employees exercise oversight of the safety programs.
- Encouraging reporting without retaliation. Employers may want to invite workers to report AI failures or concerns and make sure no one faces retaliation for doing so.
Conclusion
AI has real potential to make manufacturing workplaces safer. It can catch hazards people might miss, predict failures before they happen, and take workers away from dangerous tasks. But the technology is a safety ally, not a safety manager. Employers that pair AI with strong human oversight, clear accountability, and a solid OSHA compliance program will be in the best position to reduce injuries and manage legal risk.
Ogletree Deakins’ Artificial Intelligence and Innovation Practice Group and Workplace Safety and Health Practice Group will continue to monitor developments and provide updates on the Artificial Intelligence and Innovation, Manufacturing, and Workplace Safety and Health blogs as additional information becomes available.
Follow and Subscribe
LinkedIn | Instagram | Webinars | Podcasts
