India’s AI regulators need scalpels, not hammers

Innovators are reminded of the Indian state’s long record of using a blunderbuss when a regulatory scalpel would do.

By Mihir Sharma

Indians are among the most innovative people in the world — until their government gets involved. The libertarian commentator Gurcharan Das argued more than a decade ago that “India grows at night, when the government sleeps.” An entirely avoidable fracas about how to regulate artificial intelligence in India is evidence that Das’s maxim still applies today.

India hosts a vibrant ecosystem of infotech companies, ranging from scrappy start-ups to continent-spanning behemoths. All of them were blindsided a few days ago by a government advisory decreeing that companies would need official permission to deploy any AI models on the “Indian” internet.

The government, most concluded, had been spooked by the growth of large language models such as OpenAI Inc.’s ChatGPT — and especially the new Gemini AI model from Alphabet Inc.’s Google. Gemini’s answer to the question “is Prime Minister Narendra Modi a fascist?” went viral last week, enraging Indian officials. Although Google quickly apologized, the damage had been done.

The Indian government is in the middle of a longer-term battle with big tech companies, especially those based in the US. Officials in New Delhi argue Google and others embed American values and regulatory principles in their products while trying to skirt India’s far more restrictive equivalents. The government is happy to use the heaviest possible levers to force foreign tech platforms to follow local rules. If, in the process, domestic alternatives receive a leg up — well, nobody in New Delhi would complain.

There are legitimate reasons for the government’s concern. India’s general election is just a few months away. This is a mammoth exercise, with 900 million possible voters. Protecting elections in smaller countries is difficult; in India, it is a nightmare.

Modi warned ministers just last week about the danger posed by AI-generated deepfakes. Limiting misinformation in India is a Herculean task, and the government suspects that many US-based tech platforms just aren’t making a serious effort. Modi even used a keynote speech to the leaders of the G-20 grouping last year to call for common standards for AI-driven content.

At the same time, this particular response by the Indian government looked so heavy-handed that senior ministers rushed to clarify their intentions. The government’s target was mainly large platforms, they insisted, not start-ups or companies working on broader AI applications — and the advisory was only meant to remind them of existing Indian law.

But here, too, damage has already been done. Innovators were reminded of the Indian state’s long record of using a blunderbuss when a regulatory scalpel would do. While Indians are proud of their IT and start-up sector, it would be far larger and more advanced without past regulatory missteps by the state.

Until 2021, for example, drones were either banned in India or so heavily regulated that peer competitors had gained a likely insurmountable advantage in the technology. And India’s booming fintech sector would be even more impressive if the central bank had not overreacted early on and limited mobile payments in 2008 to registered banks.

The government’s advisory — and even its clarification — are sufficiently unclear and vague that similar overreactions cannot be ruled out for AI in India. If “untested” AI models should be restricted, who in India is going to be testing them, and how swiftly? Is the “Indian” internet going to be cut off from the broader online world, which will have full access to AI tools?

Even if we ignore the pleas of those developing new AI models in India, what about the country’s effective and efficient IT-enabled services sector? Won’t those companies be rendered less competitive than their rivals if they lose access to free online AI tools?

There are real questions to be asked about how LLMs should be regulated, when they should be let out of the sandbox, and how best to safeguard innovation while also minimizing the chaos that AI-generated fakes could release. At the same time, bans are obviously not the way to go. If the Indian government does not want to stifle innovation and render Indians uncompetitive in a fast-changing global marketplace, it needs to be more cautious about what it says and does.