Claude Training in Malaysia (2026): An HRD Corp Buyer's Guide

Only 12% of Malaysian employees say they get enough AI training at work, according to EY’s survey of 300 employees and 30 employers (EY, 2026). Most of their employers have already put money aside for it. Every company with 10 or more Malaysian staff pays 1% of monthly wages into an HRD Corp levy account (HRD Corp, 2026). The budget exists. Spending it on training that changes how people work is the hard part.

A disclosure before anything else. My agency, 2Stallions, is an HRD Corp registered training provider, and its HRD Corp claimable AI courses are in this guide. Our row in every table gets the same treatment as everyone else’s.

So the useful question for a Malaysian company is narrower than “which Claude course is best”. It is which course your levy can pay for, and whether that course teaches Claude at all. Every “best AI course in Malaysia” page I found was written by a training provider, and two of them rank their own courses first. This guide is also written by a provider, so I have kept to what HRD Corp’s own pages say and what each provider states on its own site. If you also train staff in Singapore, the funding works differently there, and my guide to Claude courses in Singapore covers SkillsFuture, and my list of AI courses in Singapore covers the other tools.

Key Takeaways

• HRD Corp pays from your own levy, up to a ceiling: RM10,500 per group per day in-house, RM1,750 per person per day on a public run.
• For a team of 10, in-house usually gets more from the levy. One employer can send at most nine people to a public run.
• Levy above RM10,000 is forfeited after 24 months without a successful claim.
• Many claimable courses never name Claude. Get the trainer to confirm the tool in writing.
• Start with Anthropic’s free courses, then pay for practice on your own work.

Last checked: 13 October 2026. I recheck the rules and prices monthly.

Which Claude training should a Malaysian company choose?

Decide the format your levy can fund first, then pick the course that puts Claude on your team’s real work.

Format comes first because HRD Corp funds in-house and public courses on different terms. In-house general courses are capped at RM10,500 per group per day (HRD Corp, 2026). Public runs are capped per person, and one employer can send only 1 to 9 people to each run (HRD Corp, 2026). Those two limits settle most of the decision.

Language is the next filter. Every Claude course I found that states a language is taught in English. Some AI courses run in Bahasa Malaysia or Mandarin, but the ones I found teach ChatGPT or Gemini. If part of your team learns better in Bahasa Malaysia, raise it before you book.

The tool is the last filter. A levy-funded course only pays off if the hands-on work happens in the tool your company actually licenses. If you run Claude, confirm the exercises use Claude.

Can we pay for Claude training with our HRD Corp levy?

Yes, if a registered training provider runs it as an approved HRD Corp Claimable Course.

The scheme is called HRD Corp Claimable Courses, code “HRD CC”, though most providers still call it SBL-Khas. You apply in eTRiS, HRD Corp’s online grant system, and only a registered training provider can deliver the course (HRD Corp, 2026). An in-house run can take 2 to 50 of your staff. A public run can take 1 to 9 from your company.

The money is your own. Employers with 10 or more Malaysian employees pay a levy of 1% of monthly wages, and those with 5 to 9 can opt in at 0.5% (HRD Corp, 2026). When you train, HRD Corp pays the provider from that balance, up to the ceiling in its Allowable Cost Matrix. You pay anything above the ceiling directly.

Unpaid levy or late-payment interest blocks the grant until it is cleared. The course also has to be registered to the provider with a programme number. A well-known trainer does not make a course claimable on their own.

There is a second route if you have someone in-house who can teach Claude. The SBL scheme funds training by internal trainers, with an allowance of RM1,400 per group per day (HRD Corp, 2026).

The national pot is large. Budget 2027 allocates RM3 billion from the levy fund for 3 million training opportunities (Bernama, 2026). What I could not find is a scheme just for AI. No HRD Corp document I could verify gives AI courses a higher ceiling in 2026, so every figure in this guide uses the general ceilings.

Is our unused HRD Corp levy about to expire?

Only the amount above RM10,000 is at risk, and only after 24 months without a successful claim.

The rule comes from Section 25 of the PSMB Act 2001 and Employers’ Circular 7/2019. Until 1 January 2020 the window was five years. It is now two (HRD Corp, 2026).

The two-year clock runs from your last successful claim, so any approved claim resets it. If you have never claimed, it runs from the levy statement date when your balance first reached RM10,000. And the first RM10,000 always stays in the account, so a balance below that is never forfeited.

A worked example, on my reading of the rule. Say your balance is RM45,000 and your last successful claim was in November 2024. From November 2026, the RM35,000 above the floor can be forfeited. The first RM10,000 stays either way.

To check your position, ask whoever manages your eTRiS account for two things: the current balance and the date of the last successful claim. If the balance is well above RM10,000 and the claim date is more than 18 months old, plan training now. An in-house course needs its grant application at least 14 days ahead, and the claim only counts once it succeeds.

Have you ever been told the levy is “use it or lose it”? That is half true. You lose only the surplus, and only after two quiet years. It still adds up for a mid-sized company.

In-house or public: which gets more from the levy?

In-house, once ten or more people from one company need training. Public runs suit smaller numbers.

The difference is in how the ceilings work. In-house general courses are capped at RM1,500 per hour per group for up to 7 hours, so RM10,500 a day for the whole group (HRD Corp, 2026). Public courses are capped at RM1,750 per person per day, the same for face-to-face and remote delivery (HRD Corp, 2025).

For a team of 10 on a two-day course, the ceilings work out like this:

The public row is the fact most course pages leave out. A team of 10 cannot all go to one public run on levy, because the scheme caps each employer at nine per run. You would split them across two runs, or pay for the tenth person yourself.

Public Claude courses also price close to the per-person ceiling. The two-day courses I found list between RM2,600 and RM3,500 per person, and at least one adds 8% SST on top (AITraining2U, 2026).

Group size has limits too. HRD Corp requires at least 2 trainees per group, prorates the fee below 5, and caps technical courses at 25 per group (HRD Corp, 2026). Remote in-house figures come from HRD Corp’s remote training FAQ (HRD Corp, 2026).

Money is not the only difference. A public class mixes companies, so the exercises stay generic. An in-house class can work on your own reports, proposals and client emails, which is where most of the value sits.

Why doesn’t our HRD Corp course mention Claude?

Many claimable courses are registered around skills, so the tool used in class is often left unnamed.

When I checked 21 Malaysian providers advertising claimable AI training on 13 October 2026, 12 mentioned Claude on the page and 9 did not. The nine mostly named ChatGPT, Copilot or Gemini, or no tool at all. Some of the twelve named Claude as one tool among four.

A course without Claude in the title is not necessarily the wrong course. A course about writing reports with AI, or automating admin work, can run on Claude, ChatGPT or Gemini. The trainer chooses on the day, often based on what the client uses. Our own courses at 2Stallions are written that way. They are tool-agnostic, and we train on whichever of the three the client runs.

The risk sits with the buyer. If your company pays for Claude seats and the trainer teaches ChatGPT, your levy has paid for skills your people cannot use on Monday. So ask the provider to confirm in writing which tool the hands-on exercises use. A one-line email is enough, and it settles the question before the grant is filed.

Seats matter for the same reason. Each trainee needs access to the tool during the course. Claude’s Team plan costs US$25 per seat per month, or US$20 per month billed annually, with a minimum of 2 seats, and Anthropic’s prices exclude applicable tax (Claude, 2026). I found no verified ringgit price.

The opposite check is worth doing too. A page with Claude in the title may be a multi-tool course. Read the outline for how many hours go to Claude itself.

Who runs Claude training in Malaysia, and what does each cost?

At least a dozen providers do. Public two-day Claude courses mostly list RM2,600 to RM3,500 per person.

Each row below repeats what the provider’s own page says, checked on 13 October 2026. “Claimable” is the provider’s wording, not my verification. Please check the programme number in eTRiS before you rely on it.

Only six of the fourteen name a trainer, ours included, and I would ask the others before booking. One provider prices its claimable seat RM300 above its self-funded seat, which is worth a question. Every Claude course here that states a language teaches in English.

Our own prices were set against the ceilings, and both are quoted before 8% SST. Ten people in-house at RM2,100 is RM21,000, exactly the two-day group ceiling. Above ten, the ceiling stays per group. Fifteen people would cost RM31,500, and RM10,500 of that would not be claimable. For twenty or more, two groups of ten each fit the ceiling. The public price of RM3,500 equals the two-day per-person ceiling, with the usual limit of nine per company per run.

What does each Claude course actually teach?

Most Claude courses teach only part of Claude. The title rarely tells you which part, so read the syllabus against the features below.

A quick guide to the terms. A Project is a workspace with your files and standing instructions for one recurring job. A skill is a saved way of doing a task that anyone on the team can run. A connector links Claude to tools such as Google Drive or email. Cowork is Claude working through a longer, multi-step task on its own. Claude Code is Anthropic’s coding agent, now used by non-developers too.

Cowork is being folded into the main Claude chat. Anthropic announced the change on 16 September 2026 and is rolling it out plan by plan (Claude, 2026). I kept the Cowork column because courses still teach the skill it names, Claude working through a longer, multi-step task on its own, and still use the name. Read it as that skill, not as a separate product.

I read the course pages of seven Claude-specific providers on 19 October 2026. “Own work” means the page says learners use their own tasks or documents, not generic exercises. Agmo’s non-technical track publishes no syllabus, so the row is its technical track. Gemrain and NobleProg publish no module list for these courses, only a short description, which is why their rows are mostly question marks. NobleProg’s page names Cowork and none of the other features. The two 2Stallions rows come from our registered HRD Corp outlines, which describe capabilities and name no product. Participants practise on whichever assistant their company approves, so Projects, Skills and Cowork apply when that assistant is Claude. I marked a feature where the outline teaches the same capability: saved instructions with reference files for Projects, and an agent that completes a multi-step task for Cowork.

The table shows a split. Three of the seven are aimed at developers or technical staff, with Claude Code at the centre. For office teams, only KelC lists Projects, and only AITraining2U, KelC and NobleProg name Cowork. Skills appear on two syllabi, and on a third provider’s product tour. Among the other providers, only AITraining2U and KelC say learners work on their own data or documents. Both 2Stallions outlines do. If your team writes reports and proposals, a Claude Code course is the wrong first step. If you are unsure which your team needs, my article on Claude Code for marketing explains where the line falls.

Does HRD Corp really pay 100% of the fee for SMEs?

No percentage subsidy applies. HRD Corp reimburses the fee from your own levy, up to the cost ceiling.

The most common funding claim on Malaysian course pages is “100% claimable”, and it is mostly harmless. It means the course fee sits at or under the HRD Corp ceiling, so a company with enough levy can claim all of it. It does not mean the government pays.

A second kind of claim is more of a problem. One provider’s funding page, dated 7 February 2026, promises a “100% SBL-KHAS subsidy” for SMEs with fewer than 200 staff and 50% to 80% for larger firms. I could not find either figure in any HRD Corp document I read. The scheme pays up to the Allowable Cost Matrix ceiling, and the money comes from the employer’s own levy balance (HRD Corp, 2026). Company size does not change the percentage.

Why does it matter if the result looks the same? Because the levy is a finite balance. A company that believes training is free tends to spend its balance on whatever course reaches it first. A company that knows it is spending its own money asks harder questions about the syllabus.

Watch the ceiling as well as the label. If a course costs more than the ceiling, the company pays the difference directly, whatever the page says about claimability. And no grant is approved while levy or interest is outstanding.

None of this makes a provider a bad trainer. Funding pages are written by sales teams, and the trainer may be excellent. Just read the funding claim against HRD Corp’s own wording before you budget on it.

How does the HRD Corp claim process work?

You apply in eTRiS before the course starts, and the provider files the claim afterwards.

The steps, in order:

1. Agree the course with a registered provider and get three documents: a quotation or invoice, the trainer’s profile and the training schedule (HRD Corp, 2026).
2. Clear any outstanding levy or interest, because either one blocks the grant.
3. Apply in eTRiS at least 14 calendar days before an in-house course. For public courses, the notice is 3 days from 15 June to 31 December 2026, and 14 days again from 1 January 2027 (HRD Corp, 2026). Earlier versions of the guidance worded the public rule differently, so confirm it with your provider.
4. If the provider asks, you can pay 1% to 30% of the fee upfront.
5. Run the training. For remote courses, the provider sends session details to HRD Corp’s inspectorate at least 24 hours ahead, and an officer may join live (HRD Corp, 2026).
6. The provider files the claim in eTRiS within 6 months of the course ending. A manager-level officer from your company signs the JD/14 form (HRD Corp, 2026).

The claim also needs the T3 attendance form, daily attendance records and an invoice made out to Pembangunan Sumber Manusia Berhad with the provider’s SST number. Your part is small, but it has a deadline. The JD/14 is the easiest step to delay, because it needs a manager-level signature.

One practical point on timing. If your levy is close to forfeiture, the clock only resets on a successful claim. Leave room for the 14-day application and the provider’s paperwork.

Is there free Claude training in Malaysia?

Yes. Anthropic’s courses are free, and two Malaysian government AI programmes are free but not Claude-specific.

Anthropic’s free learning site is Claude Academy. On 13 October 2026, its course catalogue listed 22 courses and one learning path, with completion certificates (Anthropic Courses, 2026). The titles include Claude 101, Claude Code 101, Introduction to Claude Cowork and AI Fluency for Small Businesses. For an office team, Claude 101 is the obvious start.

Two government programmes teach AI in general. AI Untuk Rakyat, from MyDIGITAL and Intel, is a free self-paced course in two modules, available in Bahasa Malaysia, English, Tamil and Mandarin. It passed 1 million completions within six months of launch (MyDIGITAL, 2024). For younger staff, the Digital Ministry’s AI for the People programme launched on 30 August 2026 for Malaysians aged 18 to 30, with six modules and a target of 100,000 people by 2027 (Bernama, 2026). Passing it gives three months of free access to tools including IlmuChat and Gemini Enterprise, but not Claude.

What do the free options miss? Your own work. Every learner gets the same examples, and no one looks at the way your team has configured Claude. That gap is what the levy should pay for.

I would make Claude 101 a condition of any paid course. It takes an afternoon, and it means the paid days start at Projects and real tasks rather than at logging in.

What should Claude training cover on PDPA and data?

What staff may paste into Claude, how a breach gets reported, and who stays responsible for the output.

A law firm shows why. Mavin Rajah, who runs Rajah & Associates, wrote after a remote session I ran with him on Claude in legal practice: “AI does not replace a lawyer’s judgment, experience or responsibility.” (LinkedIn, 2026). Almost everything a law firm handles is confidential, so the data rules come first. The same applies to HR, finance and anyone holding customer records.

The PDPA Amendment Act 2024 came into force in stages from January to June 2025. It made data processors directly liable, introduced data protection officers and mandatory breach notification, raised the maximum fine from RM300,000 to RM1 million, and replaced the old cross-border whitelist with a risk-based framework (Mayer Brown, 2025).

The guidelines add numbers your trainer should know. A breach must reach the Commissioner within 72 hours and affected people within 7 days, and a breach counts as significant at 1,000 or more data subjects. A DPO is needed above 20,000 data subjects, or 10,000 for sensitive data (DLA Piper, 2025).

Malaysia’s AI governance guidelines (AIGE) are voluntary and set seven principles, including privacy and accountability (Malaysia.gov.my, 2024). And the AI Governance Bill is coming: on 10 September 2026 the draft went to Cabinet, with tabling expected by the first quarter of 2027 at the latest (The Star, 2026).

I could not verify any way to keep Claude processing inside Malaysia. On AWS Bedrock, Claude reaches the Malaysia region only through global cross-Region inference. Treat prompts as data leaving the country. This is practical guidance, not legal advice.

Is Claude certification worth it?

For most staff, no. Only people at Claude Partner Network members can sit Anthropic’s exams.

The exams exist, but candidates must work for a member of the Claude Partner Network (Anthropic, 2026). The network is aimed at firms that bring Claude to market, such as consultancies and resellers (Anthropic, 2026). A Malaysian bank, factory or law firm that simply uses Claude is unlikely to qualify.

That changes what a “Claude certification” course in Malaysia can give you. Agmo Academy runs a two-day preparation course in Kuala Lumpur for the architect-level exam, which its page calls CCAR-F, and a non-technical track alongside it (Agmo Academy, 2026). That is useful if your company is, or will become, a partner. If it is not, your staff can prepare but cannot sit the exam.

Other courses give a certificate issued by the provider. That proves attendance, which has its place, but nobody outside your company will treat it as an Anthropic credential.

For most teams, the realistic official proof is free. Claude Academy’s courses come with completion certificates (Anthropic Courses, 2026). If you want evidence that staff have done the basics, ask for those.

I would put the levy into capability before certificates. A team that has set up Claude on its own monthly report has something a certificate cannot show. Isn’t that what you were buying in the first place?

How do we check a provider before we book?

Ask for the eTRiS programme number, confirm the course is active, and confirm the trainer is HRD Corp accredited.

Status is the check most buyers skip. A course can be listed in eTRiS and still not be claimable, because registration and approval are separate steps. Ask the provider for the training programme number, and book only once they confirm in writing that the course is active for your training dates.

Trainer accreditation is the second check. HRD Corp accredits trainers under its HRD-TDF scheme. Ask the provider to name the trainer who will actually deliver your course, then confirm that person is accredited. Only six of the fourteen providers in my table name a trainer on the page, ours included, so you will often need to ask.

Then ask the questions that decide whether the two days change anything:

1. What is the eTRiS training programme number, and is its status active today?
2. Who delivers our session, and are they HRD Corp accredited?
3. Which tool do the hands-on exercises use? Please confirm in writing.
4. Will our people work on their own tasks and documents?
5. Does the session cover PDPA and what staff should never paste in?
6. Which language is it taught in?
7. Does the quoted price include 8% SST, and does it fit under the HRD Corp ceiling?
8. Is the claimable price the same as the self-funded price?

A good provider answers all eight in one email. If the answers are vague, so is the course.

Where do I stand, and what would I buy?

Ten or more people: an in-house course on your own work. Fewer: Claude 101 first, then a public run.

I run AI training through 2Stallions, so weigh this accordingly.

My view comes from running these sessions. The gap I see most often is context: people ask AI for things without giving it the background a new colleague would need. Mavin Rajah put it better than I can in his post about our session, “AI for Lawyers: Working Smarter with Claude”. He wrote: “Simply having access to AI is not enough.” And: “The real advantage comes from knowing how to prompt it properly, providing the right context, clear instructions and defined objectives.” (LinkedIn, 2026)

That is why I would put most levy into in-house training for any team of ten or more. The ceiling covers the group, and the exercises use the team’s own documents. EY found that 81% of Malaysian generative AI users save significant time, yet 68% say their workload has risen (EY, 2026). I think training on real tasks is part of how saved time becomes lighter work.

For one to nine people, a public course from the table above is fine. Pick one that names its trainer and teaches the tool you already use. Agmo, Brainymation and AITraining2U all name their trainers. If Claude Code is the goal, Brainymation and Agmo cover it.

If your team has trained before and AI still has not made it into daily work, the gap is usually in the rollout. My getting AI past the pilot stage workshop deals with that.

If you want to talk about a levy-funded session, see Claude AI training for teams and tick the “HRD Corp-claimable course (Malaysia)” box when you enquire. I will reply with the course options that fit your team.

Frequently Asked Questions