Daybreak - Big AI is writing India’s startup rules faster than the regulator can read them

Episode Date: November 26, 2025

There's a quiet tension underlying India’s AI boom. Startups are swiftly building bold products on foundations they don’t control. From synced ride-hailing fares to the regulator with only a single office, we look at the strange mix of innovation, vulnerability, and policy catch-up shaping the space. What happens when the platform you rely on starts competing with you?Tune in.Daybreak is produced from the newsroom of The Ken, India’s first subscriber-only business news platform. Subscribe for more exclusive, deeply-reported, and analytical business stories.

Hi, this is Rohan Dharma Kumar. If you've heard any of the Ken's podcasts, you've probably heard me, my interruptions, my analogies, and my contrarian takes on most topics. And you might rightly be wondering why am I interrupting this episode too. It's for a special announcement. For the last few months, I and Sita Raman Ganeshan, my colleague and the Ken's deputy editor, have been working on an ambitious new podcast. It's called Intermission.

We want to tell the same. secret sauce stories of India's greatest companies. Stories of how they were born, how they fought to survive, how they build their organizations and culture, how they managed to innovate and thrive over decades, and most importantly, how they're poised today. To do that, Sita and I have been reading books, poring over reports, going through financial statements, digging up archives, and talking to dozens of people. And if that wasn't enough, we also decided to throw in video into.

to the mix. Yes, you heard that right. Intermission has also had to find its footing in the world of multi-camera shoots in professional studios, laborious editing, and extensive post-production. Sita and I are still reeling from the intensity of our first studio recording. Intermission launches on March 23rd. To get an alert, as soon as we release our first episode, please follow Intermission on Spotify and Apple Podcast. or subscribe to the Ken's YouTube channel. You can find all of the links at the ken.com slash I am. With that, back to your episode.

Karthi Parshotam runs a healthcare startup called One Hat, and it runs on AI. Now, I want you to take a second to guess what might his biggest worry be as of now. I bet your first thought, just like mine, was, will his patients trust an AI doctor? Turns out, surprisingly, we are quite wrong. Actually, Carthy is more worried about whether the AI will trust him. You see, his startup, One Hat, is built on an invisible layer of AI. Voice notes that turn into medical records, doctor twins that remember patient histories,

and bots that remind you to take your pills. Even AI co-pilots for insurers. None of this looks flashy on the surface. But underneath, every tool he builds depends on large language models or LLMs that Karti does not own. These models are from OpenAI, Anthropic, Google and Meta. These companies supply the plug, the power and the brain. And at any moment, any of them could raise prices, block axes, or decide to compete with him directly.

And here is where it gets scary. regulators are still not done drafting the guardrails. This is not just Karthi's problem. India's competition watchdog, the CCI or the Competition Commission of India, is looking at it too. In October, it released its first market study on AI. And its big idea, everyone should self-audit their algorithms, keep a paper trail, track data sources, inputs and objectives. A reasonable request. except for one big contradiction.

90% of Indian AI firms do not even have reliable access to data in the first place. It's like asking tenants to catalog furniture that does not even belong to them. It belongs to the landlord. Now, if you listen to Daybreak regularly, you will know because we have spoken about it many times. India's AI boom is built almost entirely on borrowed infrastructure. Nearly all of the country's 900-gen-AI startups are verticalizing, yes, but essentially they are wrapping foreign models in local skin. So what happens when the model owner decides to show up?

And that is where the real tension lies. If the foundations belong to big AI, what future is left for the people building on top of it? Welcome to Daybreak, a business podcast from the Ken. I'm your host, Nickdaa Sharma, and I don't chase the new cycle. Instead, every day of the week, my colleague Rachel Varghees and I

will come to you with one business story that is worth understanding and worth your time. Today is Thursday, the 27th of November. Two ride hailing apps side by side on your phone and watch what happens. Fairs will rise and fall in sync. Not perfectly, but close enough

to make you wonder whether the algorithms are actually whispering to each other. Competition lawyer Nitish Sharma told my colleague Indarpal Singh, a reporter with the Ken, that this is exactly the kind of thing that CCI should examine. If pricing software learns from the same data and reacts in the same ways, you can get collusion without any humans involved. It's almost like a black box conspiracy. And regulators are not used to this. They are used to chasing boardrooms, not lines of code. And that is why founders like Karti feel uneasy. When he left McKinsey in 2020,

he was not preparing to stare down Google's servers. He was just wanting to fix healthcare in India. After moving back from the US, he had seen the problem firsthand while managing doctor visits and insurance claims for his children. Access was not the issue. Continuity and accountability were. One hat was his attempt to bring order to India's healthcare chaos.

And technically, it is going well. He has tested multiple AI stacks, He has built a modular system and he can swap models when required. But the problem is accuracy. Most Indian native stacks cannot match the big players. As Karthi says, the whole world is building on top of four guys right now. And those four, OpenAI, Anthropic, Google, and maybe Meta, can shift the ground beneath him whenever they want. Google's own AI marketing tool, Pomelli, is an example.

Startups building similar. tools suddenly find themselves competing with the very platforms that they relied on. Now, many of you will know that this is an old pattern. Build on the platform until the platform builds on you. Only now, the algorithms learn faster than any cloud service ever did. Shishong K from Redacto sees it constantly. Every time OpenAI or Anthropic releases something new, about 10 companies shut down. The day you build a wrapper, he says the platform can release its own version tomorrow.

And he has seen this before. When the cloud came, people feared that big companies could just pull the plug. Eventually, though, the Indian regulatory authorities stepped in with data localization rules. Today, the actors have changed, but the fear hasn't. And there is another twist this time. You cannot find an algorithm. You cannot prosecute it. And the money is often not in the technology itself,

but in how you use it in specific domains. And that is why Karthi keeps focusing on depth. He warns one hat to be so domain specific that even if the giants enter his pace, he will still matter. And while founders fight to build their moats, the institutions meant to protect them are struggling to keep up. More on this in the next segment. Stay tuned.

I'm pausing this episode to ask you a very quick question. What's your best AI prompt? One you're actually kind of proud of. It saves you time and it does exactly what you needed to do. Why do we ask? Well, let's be honest. There's a learning curve with AI. And most of us are still figuring it out.

So, we are running a short survey to ask you to share your tried and tested hacks. You share your prompt, you get a chance to be featured. in one of our episodes, and you get to hear from others like you. All the productive, efficient folks out there who are making AI work for them. The link is in the show notes. It will take you just five minutes. We can't wait to hear from you. Now, back to the episode.

In India's antitrust circles, there is a running joke. One lawyer does half of the competition commissions work, and his name is Abir Roy. Over the last decade, Roy has filed cases against almost every moment. major tech company in India. Google, Amazon, Swiggy, Zomato, Flipkart, Apple, and the joke lands because CCI, the country's antitrust watchdog, simply has not scaled. It has one active office in New Delhi, not in New Delhi, not in Bangalore where most tech companies actually are. It's not even in Mumbai where the money flows. It's just in Delhi. Competition lawyer Nisha Kaur-Ur-Ruroy said it plainly. It just isn't business-friendly. For small players, the cost and the effort of fighting

cases shoot up instantly. The regulator's presence in other cities is either tiny or barely functional. Meanwhile, other Indian regulators like Sebi or the Aviation Authority have 20-plus offices each. And the CCI's capacity issues go even deeper. Its head count is roughly the same that it was in 2009. The Director General's office, which should have around 20 officers, has seven. The merger control team has six. Case disposals that once took three months now take twice the time. Talent is also an issue.

Uber-Roy points out that Europe's competition regulators have data scientists. India's do not. The CCI's leadership has often consisted of retired officials from other fields, auditing, policing, diplomacy. Some lawyers even joked that many fall asleep during hearings. The CCI did not respond to the Ken's questions about any of this. So, given these constraints, maybe it is no surprise that the Competition Commission's first AI market study suggested something so mild.

Keep records and audit yourself. But experts say that the study stops short. It ignores the infrastructure risks. skips the talent war, and it misses acquiring and hiring as a problem entirely. Meanwhile, founders feel the ground-shifting. The AI giants are not waiting for the new rules. They are building faster than regulators can read their own reports. Anthropic CEO has already met with ministers in Delhi.

Google is putting $15 billion into a data center in India. OpenAI made ChartGPT go-free for a year, chasing India's next billion data points. Startups feel squeezed from all sides. Raghavan from Pradesh says that getting the right data is hard, expensive and labor intensive. Compute is limited and those who've had access to it for years now have an unshakable advantage. Funding has tightened too. Investors now prefer revenue-ready startups instead of ideas. Other countries, meanwhile, are moving quite differently.

Europe writes rules before harm occurs. China has already banned illegal data acquisition and below cost pricing. India is taking a slower path. Lawyers argue that over-regulation too early might crush innovation, but under-regulation lets the giant sprint ahead. So the country watches and waits. Karthi, meanwhile, is trying to survive the middle layer, the most vulnerable part of the stack.

Big AI's new tools like Open AI's agent kit threaten companies building conversational workflows. Middleware is at the highest risk. So he is going deeper into healthcare, not broader. So now you understand what keeps Karti up at night. It is not losing to another startup. It is being replaced by the very platforms that he depends on. And as for lawyers, none of this will be real until the first live AI competition case lands.

Because when that case arrives, every theory will. will collapse. What will remain is how the algorithm behaved and whether the law can keep up. And when that they comes, founders like Karti will finally know something that they've been afraid to ask out loud. Will the algorithm still see them as partners or just as data sets that it no longer requires? Daybreak is produced from the newsroom of the Ken India's first subscriber-focused business news platform. What you're listening to is just a small a sample of a subscriber-only offerings and a full subscription offers daily, long-form feature stories, newsletters and a whole bunch of premium podcasts. To subscribe, head to the ken.com and click on

the red subscribe button on the top of the website. Today's episode was hosted and produced by my colleague, Snitha Sharma and edited by Rajiv Sien.