Ask an HSE adviser on a Malaysian upstream or downstream project what takes their time, and few will say “the field”. It is the documents: contractor HSE plans, job safety analyses, permit packs, investigation reports, monthly statistics. That is exactly the work today’s AI assistants are good at, and it is why oil and gas HSSE teams are some of the most interested people in our classes.
This guide covers where AI genuinely helps in oil and gas HSSE, where it must not be trusted, and what the industry’s own guidance says.
The pressure on oil and gas HSSE teams
Operators run on tight assurance systems. PETRONAS, for example, runs an HSE Mandatory Control Framework covering its HSE management system, contractor management and contractor audits, and NIOSH runs the NIOSH-PETRONAS Safety Passport programme for people working on its sites. Since the OSH (Amendment) Act 2022 came into force on 1 June 2024, principals also carry an explicit duty under section 18A to protect contractors, subcontractors and their workers, including information, instruction, training and supervision. The maximum penalty for breaching general duties rose to RM500,000, two years’ jail, or both.
Environmental exposure is higher too. The Environmental Quality (Amendment) Act 2024, in force since 7 July 2024, lifted maximum fines for the most serious offences, such as illegal disposal of scheduled waste, to RM10 million. Listed companies must also report work-related fatalities and lost time incident rate in their sustainability statements under Bursa Malaysia’s enhanced framework.
Six places AI earns its keep
| Task | What AI does |
|---|---|
| Contractor HSE plan review | Summarise a 60-page contractor HSE plan against your requirements and list the gaps to raise at the kick-off meeting. |
| Permit-to-work pre-read | Pull the hazards, isolations and simultaneous-operations notes for tomorrow’s permits into one page for the permit coordinator. |
| JSA first drafts | Draft a job safety analysis from the work pack and method statement, in your template, for the supervisor and HSE adviser to challenge. |
| Incident and near-miss write-ups | Turn field notes, photos and statements into a structured report and a draft timeline before the investigation meeting. |
| Lessons-learned bulletins | Rewrite an investigation finding as a one-page safety alert crews will actually read. |
| HSE statistics and reporting | Compile monthly LTIR, fatalities and training numbers for the sustainability report from the incident log. |
Typical applications taught in our AI for HSSE programme. Every output is reviewed by a competent person.
The operators are already investing
This is not fringe experimentation. In 2022 PETRONAS signed agreements with Baker Hughes, BCG and Microsoft to set up an AI centre of excellence aimed at efficiency, reliability and safety. Shell signed a global deal for Detect Technologies’ T-Pulse, which applies computer vision to camera feeds and pushes findings into its action tracking under its Goal Zero programme. In May 2026 IOGP, the producers’ association, published Report 816, guidelines for using vision AI in HSSE applications across projects and operations.
Computer vision is the headline-grabber, but for most HSSE teams the faster win is the paperwork. A contractor HSE plan review that takes a morning can take twenty minutes with a well-prompted assistant and a clear checklist.
How to start without creating new risk
- 1Pick a document-heavy taskContractor plan reviews or incident write-ups are good first candidates: high effort, low consequence if a draft is wrong, because a human reviews it.
- 2Use your company’s approved toolMicrosoft 365 Copilot or an enterprise Claude or ChatGPT plan, never a personal account. Incident data often contains personal data.
- 3Write the review rule downWho checks the AI draft, against what, and what gets recorded. Treat it like any other management-of-change.
- 4Measure and reportHours saved per report and backlog closed. That is what earns budget for the next workflow.
AI drafts; the competent person decides. In a 2024 study, GPT-4o answered 84.6% of 385 professional safety-exam questions correctly, but was weakest on emergency response and fire prevention, and small prompt changes moved its score by up to 13.5 points. That is a capable assistant, not a signatory. On a facility with major-accident hazards, the bar for review is higher still: AI never decides whether a permit is issued or a job is safe to start.
What about cameras and biometrics?
Vision AI for PPE compliance, exclusion zones and red-zone entry is spreading, and IOGP’s guidelines are the right starting point. In Malaysia, remember the privacy side. Under the PDPA amendments, biometric data such as facial recognition has been sensitive personal data since 1 April 2025, organisations must appoint a data protection officer and notify breaches, and maximum fines rose to RM1 million. Bring your DPO in before the pilot, not after. Our PDPA and AI compliance guide covers the basics.
Two days for HSSE teams on AI-assisted JSAs, contractor plan reviews, incident reports and HSE statistics, using oil and gas examples and your own templates.
Public dates coming soon — in-house classes run on your dates.
The same programme is tailored for construction, manufacturing and plantation and logistics. Teams on Microsoft 365 can also join a public Copilot class.
HRD Corp claimable. AITraining2U is an HRD Corp registered training provider. Our public and in-house courses are claimable under the HRD Corp Claimable Courses (SBL-KHAS) scheme, and we prepare the grant paperwork with you. See HRDC-claimable AI training: courses, costs and how to claim.