India AI Summit 2026: Eros Innovation wants to change how content creators interact with cinema

When you first hear Eros, chances are you might think of the media company, which for the past 30 years has been producing high-budget Bollywood films. But at the India AI Impact Summit 2026, the company came up with an altogether different game plan. 

They’ve announced that the company isn’t just producing movies anymore, but they’re also looking to explore the AI space with an innovation. The company has launched something called the large cultural model, which, as per the company, is trained on their own library of movies that they’ve produced over the past few decades. 

At the summit, I got a chance to speak with Piyush Bhatia, who is leading marketing for Eros Innovation, and had an insightful conversation with him on how the app works, the ethics of the AI, and even the copyright part of the app. Let’s take a deep dive into all I learned from him.

Also Read: Space Matrix CTO talks workplace design with AI at India AI Impact Summit 2026

How does the Eros Universe work?

First and foremost, I wanted to understand what exactly Eros Universe is and how it works. To which, Piyush explained to me that they’ve launched this app for creators, and they’re targeting it as a creator economy super app. He explained that creators will be able to use the tools to create AI-based content, from the movies that Eros has made over the past 30 years. 

Making an ‘Indian’ AI

Interestingly, rather than building on proprietary LLMs available in the market, Piyush explained that they’ve gone on the harder route of building their own, and that’s because of a few reasons. He explained that the current LLMs are essentially based on predictive text, and the problem with that is that they’re all trained on “Western data and it’s all text-based majorly.”

But he explained that with their tech, they’re trying to teach AI how to be Indian. “We are teaching it our ragas, our rasas. We are trying to teach our Vedic knowledge. We are trying to make it more Indian. We are trying to teach it cultural nuances.” 

Deep diving more into it, he said, “We have taken all of the movies that we have made in the past and we have converted it into 1.5 trillion tokens of data. And now we are training the existing open source models on this data, thereby making it more Indian. We are trying to teach it how to react to us and how to be culturally more relevant and also how to be more Indian.”

What are the benefits of using a localised model?

Once I understood the AI powering the entire thing, it was also important to understand the benefits a consumer gets from using a localised model. To which, Piyush explained it from an interesting point of view, and that had to do with the pricing. 

He said,  “Everything from voice, text, video tools are available inside and the biggest point is that it’s 50-60% cheaper than your Western models. So if you were to theoretically be paying $5 somewhere, you are going to get it for $1.5-$2. Same amount of tools, similar number of tokens. And this is happening because you are using a localized model.”

Eros on solving the copyright puzzle

When we talk about cinema and content creation, these two things don’t really go in the same picture together. A major reason behind this has to do with the strict copyright norms that media companies follow. But with the Eros Universe app, Piyush explained to me how they’re tackling this problem. 

He explained that Eros has bypassed this issue because their dataset is entirely “ethically sourced” from their own copyrighted material produced over decades. As for the ownership, Eros owns the copyright for the movies and characters used in the training as well. 

Not only that, but as per him, when a creator uses these tools to make new content, they aren’t locked into the platform. They are also free to take their AI-generated work and post it on any social media platform they choose.

What about content moderation?

Last but not least, I also wanted to understand the steps that the company is taking in order to safeguard its AI with regard to censorship. In an era of the “Grok drama” and concerns over internet safety, content moderation has become a big thing. Though, as per Piyush, Eros has taken a unique path toward this. 

Rather than relying solely on post-publication removal, they have integrated cultural sensitivity directly into the model’s training. As per him, the AI tools are trained in such a way that they can prevent the generation of culturally insensitive content. He says that this approach was developed under the guidance of the MeitY and the India AI Mission to ensure the technology remains safe and respectful.

Also Read: Young adults face big career risk due to AI, says Leverage Edu founder Akshay Chaturvedi