Introduction
Artificial Intelligence (AI) is becoming a normal part of modern workplaces. Companies are using AI for data analysis, planning, automation, monitoring, document preparation, customer service and many other activities.
AI can make work faster and easier, but like any other technology, it can also introduce new risks.
In a workplace, people may trust an AI-generated recommendation without checking it. An automated system may make an incorrect decision. Workers may become over-dependent on technology, or an AI-controlled machine may behave differently from what the operator expects.
These situations create a new area of workplace safety that organizations need to understand.
AI safety is therefore not only a technology issue. It can also become a workplace health and safety issue when AI influences machines, decisions, work processes or employee behavior.
For example, imagine a warehouse using an AI-based system to monitor material movement. If the system incorrectly identifies a safe route as available, a worker may enter an area where a vehicle is moving. The problem is not simply an "AI error." It can become a real workplace safety risk.
This article explains the emerging safety risks associated with Artificial Intelligence and how organizations can manage them.
What is Artificial Intelligence?
Artificial Intelligence (AI) refers to computer systems that can perform tasks that normally require human intelligence.
These tasks may include:
- Analyzing information.
- Recognizing patterns.
- Making predictions.
- Understanding language.
- Generating content.
- Supporting decisions.
- Controlling automated systems.
AI may be used through:
- Software applications.
- Cameras and monitoring systems.
- Robots.
- Automated vehicles.
- Predictive maintenance systems.
- Chatbots.
- Decision-support systems.
AI does not automatically mean that a workplace is unsafe. The risk depends on how the technology is designed, used and controlled.
Why are AI Safety Risks Different?
Traditional workplace hazards are often easier to recognize.
For example:
- A wet floor can be seen.
- A damaged electrical cable can be identified.
- A moving forklift can be observed.
- An unstable pallet can be inspected.
AI-related hazards can be less visible.
A system may appear to be working correctly while producing an incorrect recommendation or prediction.
This creates a new challenge:
The workplace may look safe while the technology influencing the work may contain hidden risks.
Common AI Safety Risks in the Workplace
1. Incorrect AI Decisions
AI systems can sometimes provide incorrect outputs.
For example, an AI system may recommend that a particular maintenance activity is not required because it predicts that equipment will continue operating normally.
If the prediction is wrong, equipment failure could occur.
The important point is that AI output should not automatically be treated as fact.
2. Over-Reliance on AI
One of the biggest risks is when employees start trusting AI too much.
For example:
An employee asks an AI system whether a chemical can be stored next to another chemical and receives an incorrect answer.
If the employee follows the answer without checking the applicable SDS, storage procedure or compatibility requirements, the result could be dangerous.
AI should support competent decision-making, not replace required safety checks.
3. AI Hallucination or False Information
Generative AI can sometimes produce information that sounds correct but is actually wrong.
This can be particularly important in safety-related activities.
For example, AI could generate:
- An incorrect PPE recommendation.
- An incorrect emergency procedure.
- An incorrect chemical handling instruction.
- An incorrect legal requirement.
- An incorrect calculation.
A professional should verify safety-critical information before using it.
4. Incorrect Risk Assessment
AI may be used to support HIRA or risk assessment.
It can help identify possible hazards, but an AI system may not understand the actual workplace conditions.
For example, an AI-generated HIRA may identify:
Hazard: Forklift movement.
But it may fail to consider that in the actual warehouse:
- Pedestrians share the same route.
- Visibility is poor at one corner.
- Loading activity happens nearby.
- Drivers reverse frequently.
A risk assessment should therefore include actual workplace observation and employee participation.
5. Automation and Unexpected Machine Behavior
AI is increasingly being used with automated machines and robots.
Possible risks include:
- Unexpected movement.
- Incorrect object recognition.
- Failure to identify a person.
- Incorrect navigation.
- Unexpected stopping.
- Unexpected starting.
The risk becomes more serious when people work close to AI-controlled equipment.
Appropriate safeguards, procedures, training and human oversight are important.
6. AI-Based Camera Monitoring
AI cameras may be used to detect:
- PPE compliance.
- People entering restricted areas.
- Vehicle movement.
- Unsafe behavior.
- Smoke or fire.
- Crowding.
These systems can be useful, but they may also make mistakes.
For example, a camera may fail to recognize a worker wearing safety glasses because of lighting, camera angle or image quality.
Employees should not be automatically blamed based only on an AI detection.
The system output should be reviewed where appropriate.
7. AI and Workplace Surveillance
AI can analyze employee behavior, movement and productivity.
While monitoring can support safety, excessive surveillance may create other workplace concerns.
Employees may feel:
- Constantly monitored.
- Under pressure.
- Afraid of making mistakes.
- Less comfortable reporting problems.
Safety monitoring should therefore be implemented responsibly and transparently.
8. Cybersecurity and Safety
AI systems connected to workplace equipment can also create cybersecurity risks.
A cyber incident may affect:
- Automated equipment.
- Access control.
- Alarm systems.
- Monitoring systems.
- Industrial control systems.
If a cyberattack affects a safety-related system, the consequence may not be limited to data loss.
It could potentially affect physical safety.
Therefore, cybersecurity can become part of workplace risk management where AI is connected to operational systems.
9. Loss of Human Skills
When employees depend heavily on automated systems, they may gradually lose practical skills.
For example, if an AI system always provides maintenance recommendations, workers may become less familiar with identifying early signs of equipment problems themselves.
If the system later becomes unavailable, employees may not be prepared to work safely without it.
Organizations should maintain appropriate human competence even when automation is introduced.
10. Reduced Human Attention
Automation can sometimes make employees less alert.
If a system performs most of the monitoring automatically, workers may assume that the system will identify every problem.
This can create a dangerous situation where employees stop actively observing their workplace.
Technology should improve safety awareness, not replace it.
11. AI Bias
AI systems learn from data.
If the data used to develop a system is incomplete or biased, the resulting system may perform poorly in certain situations.
For example, a safety monitoring system may work well under normal lighting but perform poorly in dark areas.
Organizations should consider whether the AI system has been tested under the actual conditions in which it will be used.
12. Poorly Designed AI Interfaces
AI systems can create risks when information is presented in a confusing way.
For example, a system may show:
"Risk Level: Low"
but fail to explain why.
A worker may trust the displayed result without understanding the limitations.
Safety-related AI systems should provide information that users can understand and act upon appropriately.
13. Emergency Response Risks
AI can support emergency detection, but organizations should not assume that AI will always detect an emergency correctly.
For example:
An AI camera may fail to identify smoke because of:
- Poor visibility.
- Camera obstruction.
- Unusual smoke characteristics.
- System malfunction.
Emergency systems should have appropriate backup arrangements.
Employees should still know how to respond manually when required.
14. AI-Generated Safety Documents
AI can help prepare:
- SOPs.
- Toolbox talks.
- Safety checklists.
- Training material.
- Risk assessment drafts.
- Safety alerts.
This can save time.
However, an AI-generated document may contain:
- Incorrect information.
- Missing controls.
- Incorrect terminology.
- Outdated requirements.
- Unrealistic procedures.
Safety documents should therefore be reviewed and approved by competent personnel before implementation.
15. New Ergonomic and Psychological Risks
AI and automation may change the nature of work.
Employees may experience:
- Increased screen time.
- Repetitive monitoring.
- Reduced control over work.
- Pressure to follow automated recommendations.
- Stress related to technology.
- Fear of being continuously evaluated.
These issues should also be considered when assessing workplace risks.
AI Safety Example – Chemical Warehouse
Consider a chemical warehouse where AI software is used to recommend storage locations.
The system receives information about:
- Product name.
- Quantity.
- Storage location.
- Warehouse capacity.
It recommends that two products can be stored close to each other.
However, the system does not have complete information about their chemical incompatibility.
If the recommendation is accepted without checking the SDS and applicable storage requirements, an unsafe storage arrangement may be created.
Correct Approach
The AI recommendation should be treated as a decision-support tool.
The responsible person should verify:
- SDS information.
- Chemical compatibility.
- Storage requirements.
- Existing warehouse procedures.
- Risk assessment.
Only after verification should the storage decision be finalized.
AI Safety Example – Forklift and Pedestrian
Imagine a warehouse using AI cameras to detect pedestrians and forklifts.
The system normally identifies people correctly.
One day, due to poor lighting, a worker is not detected.
If the forklift operator relies completely on the AI warning system, the worker could be exposed to danger.
Lesson
AI should not become the only safety barrier.
Other controls should remain in place, such as:
- Pedestrian routes.
- Traffic management.
- Speed control.
- Visual awareness.
- Trained operators.
- Warning systems.
- Workplace supervision.
AI Safety Example – Predictive Maintenance
AI can predict possible equipment failures using historical data.
Suppose an AI system predicts that a piece of equipment has a low probability of failure.
The maintenance team decides not to inspect it.
Later, the equipment develops a fault.
This example shows why:
AI predictions should support maintenance decisions, not automatically replace inspection and professional judgment.
How to Control AI Safety Risks
Organizations can manage AI-related safety risks using a practical approach.
Step 1 – Identify Where AI is Used
Prepare a simple list of AI applications used in the workplace.
For example:
- AI cameras.
- Automated vehicles.
- AI software.
- Predictive maintenance.
- Chatbots.
- Automated decision systems.
- AI-based monitoring.
Step 2 – Identify Possible Hazards
Ask:
- What can go wrong?
- Can the AI make an incorrect decision?
- Can the system fail?
- Can workers misunderstand the output?
- What happens if the system becomes unavailable?
- Can the failure create a physical safety risk?
Step 3 – Assess the Risk
Consider:
- Likelihood.
- Severity.
- Number of people exposed.
- Frequency of use.
- Level of human involvement.
AI-related risks should be included in the organization's risk assessment where relevant.
Step 4 – Maintain Human Oversight
Important safety decisions should have appropriate human review.
Employees should understand:
- What the AI system does.
- What it does not do.
- When its output must be checked.
- When the system should not be trusted.
- What to do if the system fails.
Step 5 – Verify AI Output
Safety-critical information should be checked against reliable sources.
Depending on the activity, this may include:
- Approved SOPs.
- SDS.
- Risk assessments.
- Manufacturer instructions.
- Applicable legal requirements.
- Competent-person review.
Step 6 – Provide Training
Employees using AI systems should receive suitable training.
Training should explain:
- Purpose of the system.
- Basic operation.
- System limitations.
- Common errors.
- Reporting process.
- Emergency backup arrangements.
Step 7 – Plan for AI Failure
Organizations should ask:
What will we do if the AI system stops working?
A backup process should be available for safety-critical applications.
For example:
If an AI monitoring system fails, workers should know how monitoring will continue manually.
Step 8 – Monitor Performance
AI systems should be reviewed periodically.
Check:
- False alarms.
- Missed detections.
- Incorrect recommendations.
- User complaints.
- Safety incidents.
- Near misses.
- System failures.
The organization should learn from these events.
AI Risk Management Checklist
| Check Point | Status |
|---|---|
| AI applications identified | ☐ |
| AI-related hazards identified | ☐ |
| Risk assessment completed | ☐ |
| Human oversight defined | ☐ |
| AI limitations communicated | ☐ |
| Employees trained | ☐ |
| Safety-critical information verified | ☐ |
| Backup process available | ☐ |
| System performance monitored | ☐ |
| AI-related incidents reported | ☐ |
| Near misses reviewed | ☐ |
| Corrective actions tracked | ☐ |
What Employees Should Remember
Employees should remember a simple rule:
Do not blindly trust AI when safety is involved.
Before acting on an AI recommendation, ask:
- Does this make sense?
- Is it consistent with our SOP?
- Have I checked the relevant SDS or technical information?
- Is there a competent person who should verify this?
- What could happen if the AI recommendation is wrong?
A few seconds of verification can prevent a serious mistake.
AI and Near Miss Reporting
AI-related near misses should also be reported.
Examples include:
- AI fails to detect a pedestrian.
- Automated equipment moves unexpectedly.
- AI gives an incorrect safety recommendation.
- AI camera generates repeated false alarms.
- Workers misunderstand an automated warning.
- AI monitoring system stops working.
These events can help organizations improve both technology and workplace controls.
Safety Leadership in the Age of AI
Management has an important role in introducing AI safely.
Management should:
- Understand the risks.
- Involve workers.
- Provide appropriate training.
- Avoid unnecessary dependence on AI.
- Maintain human decision-making.
- Review system performance.
- Provide resources for safe implementation.
Introducing new technology without considering workplace safety can create new hazards instead of solving old ones.
Employee Participation
Employees should be involved when AI technology is introduced.
Workers can provide practical information about:
- Actual work conditions.
- Common equipment problems.
- Difficult operating situations.
- System limitations.
- False alarms.
- Unsafe recommendations.
The people who actually use the system every day may notice problems that are not visible during software testing.
AI Safety and ISO 45001
AI-related risks can become relevant to an organization's Occupational Health and Safety Management System where AI affects workplace activities or hazards.
Organizations may consider AI during:
- Hazard identification.
- Risk assessment.
- Operational planning.
- Change management.
- Worker consultation.
- Competence and training.
- Emergency preparedness.
- Incident investigation.
- Corrective action.
- Continual improvement.
When new technology changes the way work is performed, the related safety risks should be reviewed rather than assuming that existing controls will automatically remain effective.
Common Mistakes When Using AI for Safety
Organizations and employees should avoid:
- Copying AI-generated safety information without verification.
- Using AI as the only safety control.
- Assuming AI is always correct.
- Ignoring system limitations.
- Failing to train users.
- Not having a backup process.
- Ignoring AI-related near misses.
- Allowing AI to replace competent safety judgment.
- Using outdated AI-generated information.
- Treating AI-generated documents as automatically approved documents.
Do's
✔ Use AI to support safety improvement.
✔ Verify safety-critical information.
✔ Maintain human oversight.
✔ Train employees properly.
✔ Understand system limitations.
✔ Report AI-related failures and near misses.
✔ Maintain backup arrangements.
✔ Review AI performance regularly.
✔ Include relevant AI risks in risk assessments.
Don'ts
❌ Do not blindly trust AI.
❌ Do not use AI as the only safety barrier.
❌ Do not copy AI-generated procedures without review.
❌ Do not ignore incorrect AI recommendations.
❌ Do not allow untrained employees to operate safety-critical AI systems.
❌ Do not assume automation removes all workplace risks.
❌ Do not ignore the human and psychological effects of technology.
The Future of Workplace Safety and AI
AI will probably become more common in industrial workplaces.
It may help organizations with:
- Predictive maintenance.
- Hazard detection.
- Safety monitoring.
- Training.
- Data analysis.
- Emergency detection.
- Process automation.
The goal should not be to choose between humans or AI.
The better approach is to combine the strengths of both.
AI can process large amounts of information quickly, while experienced workers can understand real workplace conditions, unusual situations and practical consequences.
The safest approach is therefore:
AI + Human Experience + Safety Controls = Better Safety
Key Learning
The main lessons from AI safety are:
- AI is a tool, not a replacement for safety responsibility.
- AI can make mistakes.
- Human verification remains important.
- Safety-critical information should be checked.
- AI-related hazards should be assessed.
- Employees should understand AI limitations.
- Backup arrangements are important.
- AI-related near misses should be reported.
- Technology should strengthen existing safety controls.
- New technology should be introduced with proper planning.
Frequently Asked Questions (FAQ)
Q1. Can Artificial Intelligence create workplace safety risks?
Yes. AI can introduce risks through incorrect decisions, system failures, automation, over-reliance, surveillance, cybersecurity issues and unexpected machine behavior.
Q2. Can AI replace a safety professional?
AI can support safety professionals by analyzing information and preparing drafts, but it should not automatically replace competent professional judgment, workplace observation or responsibility.
Q3. Can AI be used for HIRA?
Yes, AI can help generate ideas and identify potential hazards, but the final HIRA should be based on actual workplace conditions and reviewed by competent personnel.
Q4. Should AI-generated safety documents be used directly?
No. Safety-critical documents should be reviewed and approved through the organization's established document-control process before use.
Q5. What should an employee do if AI gives an unsafe recommendation?
Do not follow it blindly. Stop and verify the information using approved procedures, SDS, manufacturer information or a competent person, depending on the situation.
Q6. Is AI a safety hazard by itself?
Not necessarily. AI is a technology. The risk depends on how it is designed, implemented, controlled and used in the workplace.
Conclusion
Artificial Intelligence is changing the way people work, and it can provide many useful opportunities for improving workplace safety. At the same time, it introduces risks that traditional safety systems may not always consider.
The biggest mistake would be to assume that an AI system is always correct.
A workplace should continue to use basic safety principles even when advanced technology is introduced:
Identify the hazard → Assess the risk → Apply controls → Maintain human oversight → Monitor → Improve.
AI can help us process information faster, identify patterns and support decisions. But workplace safety still depends on people understanding the work, recognizing hazards and taking responsible action.
The future of industrial safety should not be about replacing people with technology. It should be about using technology responsibly while keeping human judgment, practical experience and effective safety controls at the center of the workplace.
Use AI to support safety, but never let AI replace safety responsibility.
Discussion
Is Artificial Intelligence already being used in your workplace? What safety risks or benefits have you observed? Share your experience in the comments so other safety professionals can learn from real workplace situations.
About the Author
Written by Mahesh Chand
Warehouse Safety Professional | Chemical Warehousing | Fire Safety | ISO 9001 & ISO 45001
Through Trading Hatke, Mahesh Chand shares practical workplace safety knowledge, warehouse safety practices, industrial safety guidance and ISO-related learning in simple and easy-to-understand language.
📌 Follow Trading Hatke for more practical safety guides, warehouse management tips and ISO best practices.
Disclaimer: This article is intended for educational and informational purposes only. AI-generated information should not be treated as a substitute for competent professional judgment, approved workplace procedures, Safety Data Sheets, manufacturer instructions or applicable legal requirements.
AI in Workplace Safety – How Artificial Intelligence Can Improve Worker Safety
https://www.tradinghatke.com/2026/09/ai-in-workplace-safety-how-artificial.html
Complete Industrial Safety Guide – Practical Workplace Safety
https://tradinghatke.blogspot.com/2026/09/complete-industrial-safety-guide.html
Workplace Safety Tips – Complete Guide for a Safer Workplace
https://tradinghatke.blogspot.com/2026/09/workplace-safety-tips-complete-guide.html
Contractor Safety – Complete Guide for Workplace Safety
https://tradinghatke.blogspot.com/2026/08/contractor-safety-complete-guide-for.html
%20Overview%20Infographic.png)
No comments:
Post a Comment