New Delhi: A United States government export control directive that forced AI company Anthropic to abruptly pull access to its Fable 5 and Mythos 5 models last week triggered an unusually bitter public argument between two prominent Indian commentators. The two are often bracketed together on the ideological right—Hindutva ideologue and author Rajiv Malhotra, and former Infosys CFO and policy advisor Mohandas Pai.
The two traded insults on X soon after Anthropic’s announcement on 13 June.
“Made millions by renting Indian brains for building US intellectual property,” Malhotra wrote, accusing Pai and fellow industrialists of a “quick buck mentality” that left China with a different strategy to protect sovereignty.
Pai shot back within hours: “So says a big, big failure and economic refugee who did nothing useful in tech in his life and now points fingers at others.”
To which Malhotra responded with “a mere bean counter/admin who exploited India’s tech youth with wage arbitrage to enrich yourself.”
Pai’s final volley was blunter still: “You are a loser who ran away as you failed to scale up as a founder and gave up… Buzz off.”
The spat spilt into two separate interviews given to ThePrint afterwards, which bared a deeper disagreement over India’s AI strategy, the role of the IT services industry in it, and who is responsible for the country falling behind on large language models (LLMs).
The Anthropic directive
Anthropic disabled Fable 5 and Mythos 5 worldwide on the evening of 12 June after the US Commerce Department issued a directive ordering the company to suspend access for foreign nationals, including its own non-citizen employees.
Given the breadth of the order, Anthropic said in an X post on 13 June that it had no way to comply selectively and disabled both models for all customers globally, while other Claude models remained unaffected.
The company added that the directive, received at 5:21 pm Eastern Time, did not spell out the specific national security concern, but that it understood the trigger to be a method of bypassing Fable 5’s safeguards. Anthropic maintained that the underlying jailbreak surfaced only a handful of previously known, minor vulnerabilities that other publicly available models could find too, and that there had been no evidence of any harmful outcome from it.
Commerce Secretary Howard Lutnick had written to Anthropic CEO Dario Amodei laying out the restriction, according to the Wall Street Journal.
Also Read: AI is now building itself, getting out of Human control. Anthropic wants everyone to pause
Malhotra’s claim
Speaking to ThePrint, Malhotra used the Anthropic episode as proof of a thesis he says he has held since his 2017 book Artificial Intelligence and the Future of Power. Malhotra argued that the absence of Indian-owned LLMs could amount to a new AI-driven colonisation, with the US and China holding AI soft power, which could lead to them playing the same role as Britain and France during the Industrial Revolution.
“In the book, I have chapters like the return of the East India Company, which is what I’m referring to, some of these big tech people. The East India Company also came as a private company, and then it took over and ran the world. We have Google Devta, Facebook Devta, because we are going gaga over these as a sort of ultimate authority about information, knowledge, and almost everything,” he said.
Malhotra added that he was still using one of the suspended models when access vanished without warning, calling it “a living example” of how exposed Indian users and companies were.
“There was no warning, there was no notice, they didn’t even tell you what they were doing. They just withdrew the whole thing. And now they’re going to reintroduce it in a very careful way—monitoring who uses what and what for,” he said.
On the question of data sovereignty and the need to build a data centre in India, Malhotra was blunt in his response: “Location of the data centre has nothing to do with who owns the data. The data centre could be located in Delhi, Bombay, New York, Moon, Mars, anywhere. The question is who owns that data and who has access to the data, who has the encryption codes and who has the architecture so that you can make sense of the data. All that is in the hands of foreign people.”
But he was dismissive of Pai’s claim that India ranks among the world’s largest AI powers, comparing Indian information technology outsourcing-driven growth to colonial-era cheap labour. Malhotra said that it was an echo of the “rented Indian brains” charge he had already made on X.
“The Indian does not own a single equity in even one line of code. He’s just a labourer,” the author of Breaking India added.
For this, he blamed India’s IT billionaires and successive governments—both Congress and BJP—for treating outsourcing profits as an end in itself rather than reinvesting in indigenous chips, operating systems and search engines.
“They had so much money at the time. There was no shortage of money. There was a shortage of vision. There was a shortage of nation-building,” he said.
Malhotra contrasted India’s approach with that of China’s two-decade build-out of rare earths, semiconductors and AI infrastructure, saying Beijing had “captured all the supply chains.”
“About 20 years ago, China identified a dozen technologies of the future. They identified drones, robotics, AI, solar and so on. And they built a huge ecosystem. They captured all the supply chains. They went and bought all the rare earths. So they had vision,” he said.
But in India, Malhotra argued people were “just looking at what can I do in the next six months to make more money, raise share prices, get a few more billion dollars in my pocket.”
He also said that he had been denied a platform to make this case at an AI summit. Claiming it as further evidence of a broader unwillingness in Indian policy circles to even entertain a contrarian view.
“The biggest problem is a lack of free speech on policy debates,” he said.
Also Read: The AI jobs crisis no one is talking about
Pai’s retort
Pai’s response was personal as he dismissed Malhotra as an “economic refugee” with no track record of his own. He further accused him of “abusing others who built his big industries, created huge jobs.”
In his interview with ThePrint, though, he engaged with the substance of the charge, arguing the suggestion that Indian IT majors such as TCS, which he said brings in over $245 billion in revenue and employs roughly six million people, had “failed” by not building their own LLMs reflected a basic misunderstanding of what different kinds of companies do.
“These nuts don’t understand that AI is very different from the services business. You’re like saying the State Bank of India should invest, become a venture capital fund and invest in AI and become like Goldman Sachs. The State Bank of India is a commercial bank. That’s their business. They’re doing very well. Goldman Sachs is the investment bank. And people like Sequoia are a venture fund. A venture fund cannot run a bank. A venture fund is an investor. They all do separate businesses,” he said.
Pai went on to describe his five layers of the AI stack—power, chips, hyper-scale cloud, foundational LLMs, and applications. He argued that India currently has meaningful capacity only in the first and the last.
Power, he said, is not a constraint; chips and hyper-scale cloud are years away without a far larger investment.
The chartered accountant added that India is a cheap industry making chips of 22 and 25 nanometers, essentially just gift packaging—buying the chips and packaging is for electronics. He said that India’s chip manufacturing ability still had a while to go.
Frontier LLMs, which have cost American and Chinese players tens of billions of dollars each, are simply out of reach for now, Pai argued, given the scale of capital involved, considering the roughly $4 trillion of global investment already committed.
“The world is committed to $4trillion, and people say that it will go to $7 trillion by 2030. There is no business to justify that $4 trillion of investment. It’s a bubble that is happening. People are pouring money into the market because it is giving them a valuation. And we don’t have that industry. America has got 40 gigawatts of cloud. India has got two. We’ll go to maybe 7 or 10 by 2030,” the Padma Shri awardee said.
Pai’s reading of the Anthropic suspension was sharply different from Malhotra’s. He described Mythos as a model “based on cybersecurity” that “can get into your system and point out vulnerabilities” and argued that the US government denying India access to it raises a cybersecurity concern of its own.
“That means the CIA can use it, the NSA can use it and hack into systems and destroy the systems. That is the real threat,” he added.
“We need Indian sovereign data, sovereign hyper-scale,” Pai told ThePrint, insisting data generated in India should stay under Indian control and subject to Indian law. He claimed that the Reserve Bank of India had already formed a group with the finance ministry on the issue.
Contrary to Malhotra, Pai pushed back on the idea that Indian IT firms were sitting idle, saying they were working on hundreds of AI-adoption projects for global clients and building vertical, domain-specific models in financial services and healthcare, which, according to him, is a different business from building horizontal foundation models.
But for Pai, India’s hiring and attrition figures have not yet seen the scale of AI-linked layoffs reported in the US.
“So far, we have not seen much impact in India,” he said, though he acknowledged entry-level hiring by IT services firms has slowed even as domestic Global Capability Centres continue to hire.
Pai also urged for a larger, faster-moving government AI fund, citing that the existing scheme of roughly Rs 10,000 crore, which he said has been slow to disburse.
“Government is too slow,” he said, arguing it should be routed through institutions such as the State Bank of India rather than run departmentally, with private capital crowded in alongside it.
Both men agree, in essence, that India lacks the compute, chip manufacturing and capital base to build frontier AI models on its own terms in the near future.
But they part ways on what follows from that: Malhotra reads it as a story of national failure and elite complicity with a recolonising West. While Pai reads it as a structural, capital-intensive problem, in which Indian IT services firms are already playing the role suited to their scale in helping the world’s largest enterprises adopt AI rather than build their own. The government, in Pai’s view, needs to move faster and bigger on funding the layers India can realistically compete in.
The Anthropic suspension, which briefly cut off paying customers worldwide with no advance notice, ended up serving as the spark for a debate and a personal feud that was already simmering between the two men long before last week.
(Edited by Insha Jalil Waziri)