Glad that @NDTVProfit invited me and asked questions on what AI governance and public-private partnership trends one can fathom from the Pentagon vs Anthropic situation. Here’s my complete take on this. While the Trump admin cannot be relied upon to create policy precedents - the incident created a policy precedent: LLMs remain unreliable.

India was not fooled, this is what I told @AKLKO1977. We wanted that theatre. Problem is the development partnership side of Quad had no original ideas from India - it was literally Biden admin’s nasty opposition supported by India-US commentators in high places. The “handle against China” part was truly too much for us to handle, I agree there, and that was because it was not consistent from that one Biden admin. I still think Quad is just IND-JAP-AUS trilateral 1.

Brilliant. Hopefully these girls will upgrade the quality and talent in the officer corps, over time. Quick time - shd have many more girls recruited in officer ranks..

RT @IndusThink: The only reason I find the export debate useless is simply because these models are not technically advanced. The "export c…

ET should be trolled for this. This was published 3 days ago. Shameful.

Indian cos don’t face this problem. No one is going to take on well entrenched promoters. Indian govt (via large state owned investors) and regulators frown upon hostile takeovers or activist investors. Many promoters aren’t averse to siphoning off funds & screwing investors.

This is wrong. In US, big shareholders will exit and hostile takeovers will ensue if R&D obsession leads to under performance. Nvidia and Jensen faced this problem 12-13 years back. He didn’t back off though. In fact this was the big criticism of US shareholder capitalism model.

Even CEA believes that AI valuations resemble a bubble. Very true and correct.

You are being more funny than Kushal.

Strategic interest divergence started under Biden’s term but it became more mainstream in both apparent and illusory ways. Everyone countered the illusory way too late (godliness of the relationship by a section of commentariat as @d_extrovert points out) while the apparent way was hid. I never found people-to-people ties as the driver of this relationship that inclusively. It is what it is. Germany has similar kinds of ties with us, but we all know what former Foreign minister Baerbock has done with us. The apparent way was limiting to defence relationships. But then India focuses more on Israel and UAE (makes sense) and would still keep aspects of defence partnership there. Push forward to 2025. Trump goes berserk like totally. The apparent way is kinda more mainstream.

Soon people will start discovering that the TCS senior business analyst is cheaper than AI credits once the Anthropics of the world remove the subsidies for API access

Wow, we have a proper visual proof of people in US being absolutely bonkers. Courtesy: @MattooShashank

The only reason I find the export debate useless is simply because these models are not technically advanced. The "export control" debate being proposed by tech policy people comes from this understanding that these models have an economic value. I am saying we can revisit these things faster as a society first - government can come later - that: 1. What is the real economics of resources and tech value around these models? NASDAQ is crazy and the way tech stocks are going up to and fro - it is better to stay away from US market's frenzies. How can a shoe company suddenly claim it is an AI company next day and stock price goes up then? Are retail investors crazy? 2. Most Indian VCs and investors and even government are not aware of data points like discussed in (1). If they won't be able to understand this - they won't realise what works as a reliable AI business model. So please - don't overreact on US businesses and AI labs. Study the 3. On frontier models - we have the talent in India and can bring some back from the US. But what we actually need in addition, is we start focusing on alternative AI research too, go through the evolution curve of AI again through private focused groups - because the government is not expected to know about evolution of AI. You cannot expect ministries to be aa capable as say a Roger Penrose or anyone. So it's fine. Sarvam is a beneficiary of private funding and government support, both. We can have government funding these initiatives as @mnwsth and others who are authors of UP.AIACT.IN Report 2026 had suggested. The Gujarat model is the best there undoubtedly. Apart from this, echoing @kingofknowwhere, fund that 5L per month to researchers in India at first so that they don't feel insecure and can engage into fundamental AI research. This is urgently needed. There is no other way. Beyond this - I am not sure if you need government support. You can have support on exhibitions and all, which is normal. 4. Last, try to solve economics of open source AI and associated intellectual property problems. This hits at the stomach of every AI lab or startup or translational research lab. 5. One more thing - our researchers have to address this problem of finding alternative research streams of AI. Go through your ArxiV, Openreview and other sites and figure out what trajectory of AI Innovation and subbranches are being made. That makes sure you are not a victim of the hype and you start publishing impactful research sooner followed by concentrating more focus into frontier AI research, beyond and within LLMs. I think we can try beyond LLMs, and hence we should not make the same mistake the Americans made. Trust me - unless you address these issues - you cannot go ahead with fundamental AI research which is ambitious, focused and not scared. Hope my suggestions reach the right people. I have nothing against anyone personally but I will call out BS. Do whatever you can with this information. cc @soma_as_moon7 @RupakChatto @adityajakki @prasannavishy @AKLKO1977 @sreemoytalukdar

I know

Export controls used to mean arms, then chips. Now it’s even a chatbot. And “foreign national” is half the talent that runs Silicon Valley. China has leading open-source models. India barely has a model. Now sit with that. This isn’t a desi VC problem. It’s a sovereign crisis.

Big 4 Consultants can make even the most gifted/naturally intelligent person hallucinate with their inane presentations. What chance does poor Artificial Intelligence have?

We know exactly what your “Trusted corridor” even means. Client side snooping, controls, kill switch and access to Indian systems and data at will for US Gov for the benevolence of access to SOTA. Oh please! Can’t imagine that a 3-star retired Army General retweeted this!!

Newfound respect for MSMEs in India, especially manufacturing ones dealing with incompetent state bureaucrats. They use alternative lending mechanisms and strive to displace China despite the odds. Unlike volume hiring giants, they aren't crybabies and work without expectations.

RT @HarveenChadha: 2 labs are not enough for a country like India, we need more labs, more gpus, more research/infra engineers, more collab…

The closest that there has been to such an indication is the claim that the jailbreak research was shared with "security researchers." But there is no indication that this had influence on decision-making the 1 independent researcher who has been quoted concurs with Anthropic

The admin is going to keep leaking all sorts of details that make it sound, to a casual reader, like this was a reasonable decision. But there has been zero indicating that domain experts at CAISI or NSA were involved - all reporting points to "senior White House officials."