Engineering judgement can use AI. It cannot hand over the signature.
Design checks, inspections and plant control increasingly run through models. South African law already expects a registered person to take responsibility for engineering work. AIC makes that responsibility visible when part of the work is done by AI.
AI register: engineering
Sample data
- Code-check assistantStructural designN. Dube, Pr EngComplete
- Drone crack detectionBridge inspectionNot namedNo owner
- Pump failure forecastMaintenanceK. Naidoo, Pr Tech EngNo risk recorded
- Site safety camerasWorker monitoringSafety officerDecides about people
For
- Consulting engineers
- Asset owners and operators
- Plant and process engineering teams
- Health and safety officers
Where AI decides, and what goes wrong
Decides about: Clients and the public
Checking designs against codes
A model passes a design outside what it was trained to judge, and the check is taken as the engineer's.
Decides about: Operators and the public
Finding defects in inspections
A crack or corrosion missed in drone or camera footage because the model was confident.
Decides about: Workers and the public
Deciding when to maintain
Maintenance deferred on a forecast that was wrong in a way nobody reviewed.
Decides about: Employees
Monitoring workers for safety
A camera's flag becomes a disciplinary decision without a person looking.
A model passes a design check
The question an inspector, a client or an inquiry will ask is simple: who signed off, and what did they see?
- 1
The check runs
A structural design is run through an AI code-checking tool, which reports that it complies.
- 2
The engineer is named
The register names the registered person who answers for the tool's use on this project.
- 3
The review is recorded
The tool's output and the engineer's own review are logged together, including anything the engineer overrode.
- 4
The file stands up
An evidence pack gives the client or an inspector a fingerprinted copy of the record they can check matches the record AIC holds.
Rules that already apply
A summary, not legal advice about your organisation.
Engineering Profession Act, 2000
Work identified by ECSA as engineering work is reserved for registered persons, or work under their supervision, under the Identification of Engineering Work rules.
What it asks of youA registered person who takes responsibility for the work, including the parts an AI tool performed.
Occupational Health and Safety Act, 1993
Section 8 requires employers to provide and maintain systems of work that are, as far as reasonably practicable, safe and without risk to health.
What it asks of youEvidence that automated control and monitoring systems are part of a safe system of work, with someone accountable for them.
POPIA section 71
Applies when AI monitoring of workers leads to decisions about them, such as discipline. Read the guide
What it asks of youA person able to intervene, and a way for the employee to be heard.
EU AI Act
AI that is a safety component of machinery is high-risk from 2 August 2028. AI that is a safety component in managing critical infrastructure, such as water, gas, heating or electricity supply, is high-risk from 2 December 2027. Read the guide
What it asks of youRisk management, logging and human oversight for systems placed on the EU market or used there.
What the platform gives you
Every AI tool, with its engineer
Design, inspection and control tools on the register, each with the registered person who answers for it.
How it worksOutput and review, side by side
Log what the tool concluded and what the engineer decided, through the API or by hand.
How it worksSafety risks tied to the system
Missed defects, drift and unannounced vendor changes on the risk register, against the tool.
How it worksA file for the client or inspector
A sealed pack of the record for a project or a period.
How it works
Process industries
How AI maps onto the safety and operational frameworks process plants already run.
Questions
Does using an AI design or inspection tool change who is responsible?
No. The registered person who takes responsibility for the work remains responsible for it, whatever tools were used. AIC records that responsibility and the review behind it.
Our AI controls plant, not people. Does AIC still apply?
AIC's standard is about decisions with consequences for people, and plant control affects the safety of workers and the public. The platform's register, risks and decision log work the same way for safety-critical systems.
How does this relate to process safety management?
It sits alongside it. Process safety asks whether the plant is safe; AIC records who answers for each automated decision that affects it, and keeps the evidence.