Taxing AI Necessary To Compensate Displaced Workers

Chennai: Deployment of AI will see the transfer of massive wealth from the labour side to the capital side. Several experts, including former RBI Governor Raghuram Rajan, have suggested taxing companies using AI. This can be used to build a safety net for the retrenched workers.

However, India has already given 21 years of tax holiday for companies using services of Indian data centres. This creates a dichotomy in terms of taxing AI, finds J B Mohapatra, former chairman of Central Board of Direct Taxes.

Recently, Raghuram Rajan said that tax should be levied on the usage of artificial intelligence by corporates. How does such taxation become necessary for countries in an AI economy?

The concept of taxing AI has come from the jurisdiction where AI started — the US. The voices calling for AI to be taxed are coming from the democratic side in the US. There is also a congressional Bill, moved in June 2026, on how AI needs to be taxed.

The divide between the capital and labour sides of economics is now slowly getting exposed. Daron Acemoglu, one of the Nobel laureates, has said that the time has come to equalise taxing priorities and taxing principles between labour and capital.

AI is a mode of transferring massive wealth from the labour side to the capital side. All the voices that we will keep hearing about taxing AI are basically about building a safety net for displaced or retrenched workers who are badly affected.

How would such a tax work?

It could be a token tax where the user of AI is taxed on the basis of the unemployment it creates. The hyperscalers in the market — Oracle, Meta, Amazon and others — the big five — are the ones monopolising the AI market. The point is how to tax them.

The US congressional Bill briefly provides for this. The taxes generated would be put into a trust fund under the Treasury. From the trust fund, the money would go to the Department of Labour, where there would be a grant system. Companies that retrain workers and make them more proficient in AI processes would be given some refundable tax credit.

Is India ready for this kind of taxation?

That is the question. Through the Finance Act 2026, India has allowed foreign companies using the data centre services of an Indian company a 100% tax holiday up to 2047. So India has already given them a tax rebate for as long as 21 years.

On the user side, where Indian users are concerned, taxing them through some kind of levy has to be reconciled. You would not be taxing production but consumption. We cannot be unaware of this dichotomy.

There is a massive overdrive to bring AI in. Automation will help raise productivity. We have to tread very carefully, but the immediate concern is the security of workers who are being displaced.

Can India look at some existing models of taxation to address this?

Prior to GST, if you imported technology from outside India, there was an R&D cess of 5%. That cess went to the Technology Development Board, which used the money to improve and develop India-centric technologies for Indian consumers.

We can have those kinds of models of taxation. We must do it very modestly to begin with; otherwise, it could retard the whole growth of AI.

If an AI company develops technology in one country, holds the technology in another and earns revenue from users across the world, which country should have the right to tax those profits? Will traditional taxation based on physical presence become obsolete?

At this point in time, the permanent establishment, or PE — the fixed place of business of a company — determines its taxability. A sovereign has the right to tax the amount derived from that particular business.

There is no unanimity on including digital presence as a PE for taxation purposes. As long as the model Double Taxation Avoidance Agreement is not redrafted to include digital presence as an indicator or parameter of PE, the restrictive meaning of PE available now will determine taxability.

Should we look at other ways to tax AI? What exactly should be the criteria for taxation?

Coincidentally, all the big digital companies are American. Pillar One is already on the table at the OECD. The US has withdrawn from the Pillar One deliberations. There is a paper containing ideas on how residual tax in excess of 10% of the profit of multinational enterprises with revenues of more than $20 billion in a particular fiscal year could be transferred back to market economies where their digital presence exists.

India has also withdrawn taxes that it had imposed, such as the 2% tax on e-commerce operators and the 6% tax on Google advertisers. Whatever manner in which we imposed and subsequently withdrew digital service tax, it was not perfect and generated very little tax revenue.

Global technology companies have strongly resisted unilateral digital taxes. India has also withdrawn its tax. What are the major issues with digital tax?

Digital tax is ad hoc. It anchors taxes on revenues, not profit. So if a company has a very thin margin or is making a loss, it equally suffers.

There are also implementation challenges. And all the big technology companies are US-based. Putting a digital services tax on US companies triggers reactions from the US, where it is habitually taken as a retaliatory tariff measure, followed by a counter-measure.

There can be very strong retaliation through the Trade Act. So there are challenges in imposing digital services tax in India and elsewhere in the world.

The OECD's global tax framework sought to redistribute taxing rights to market economies. Why has achieving global consensus proved difficult?

There are two pillars. On Pillar Two, there is a general consensus. One hundred and forty countries have agreed to Pillar Two, which is the global minimum tax of 15% effective tax. Fifty-five of those 140 countries have already legislated it.

India has not yet legislated it, although it has agreed with regard to Pillar Two. The main reason could be that not many large multinational enterprises are headquartered in India. They are probably headquartered in Singapore, the Netherlands or the UK.

Pillar One was very ideally planned. The largest global multinational enterprises, those with revenues in excess of €20 billion and profitability in excess of 10%, would have 25% of a portion of their profits distributed among the market economies where those companies have a footprint.

Some part of the taxes would thus be redistributed among market economies. But that has not worked because the US, under Trump, has backed out of Pillar One.

Will national AI taxes create trade issues and become another source of tension between countries and the US?

Multilateral arrangements for information sharing are important. Under AEOI, for example, the Common Reporting Standards have more than 100 countries as partners. With regard to crypto, we have the Crypto-Asset Reporting Framework, or CARF, under which many jurisdictions have come together to exchange information on crypto assets.

But at the end of the day, bilateral agreements in the form of Double Taxation Agreements are vital in building the taxing rights of the counterpart nations deliberating on a particular taxing right.

The OECD's first-year result for 2024 says it managed to increase the effective tax rate by 1.7% and could generate additional taxes of $90 billion to $134 billion. That would be roughly 2.3% to 3.3% of corporate income tax globally.

There are voices saying these things could also have been done domestically by imposing minimum taxes at the domestic level instead of having all this paraphernalia. Multilateral arrangements for sharing taxing rights are one thing, but more important is concentrating on bilateral arrangements and agreements and ensuring that matters are mutually settled between two nations.

If every country starts imposing its own digital and AI taxes, could we end up with a fragmented global tax system and competition in taxation?

One must understand the scale of investment. Taxes will come at the end of the day, and direct taxes will come only when there is a profit. We really do not know, given the kind of capital and recurring investment in AI, how soon or how late these investments will break even and start generating profits.

Secondly, if it is an indirect tax on hyperscalers, there is no reason why they would not pass it on to consumers.

So, without knowing the scale of investment in AI, it is very difficult at this point to comprehend whether there will be competing national claims on AI taxes. It may happen, but for lack of data, we cannot comment at this point in time.

Companies investing in AI may take longer to break even, but during this transition a lot of people could lose their jobs. How will governments compensate workers displaced by AI?

The loss of jobs and employment is no longer an apprehension; it will be a reality.

The Ministry of Labour and Employment should work out sector-wise profiling of jobs being displaced and develop creative ideas to manage the employment scenario. For example, some jobs could be classified as irreplaceable, meaning they cannot be replaced through AI.

A lot of deliberation is required. At the end of the day, whatever tax optimisation happens for companies through the use of AI resources will conversely result in more taxes.

The incremental taxes will arise because companies will get the same volume of work without the manpower and payroll costs. The man is gone; you have a bot in his place or some AI resources. So, whatever costs you maximise or optimise, to that extent you should be paying more taxes.

There could be an administrative mechanism to segregate these incremental taxes and keep them aside as a corpus to be used for retraining retrenched people.

Tax is not the panacea at this point. Tax is not the solution. Tax could be one part of a solution, but it is not the solution.

You could have regulations under which some jobs are classified as irreplaceable. At the same time, you need to build a financial buffer. The congressional Bill in the US has already proposed a trust fund for the protection of workers. The trust fund would be held in the Treasury.

We can have a similar trust fund in the Public Account in India, with an administrator for the protection of workers. In the US, that would be the Department of Labour; here, it could be the Ministry of Labour.

That administrator could manage the trust fund for training, retraining and rehabilitating retrenched workers. There are many ways it can be done, but it requires longer and larger deliberation.