The last few years have seen an exponential growth in the global tech landscape. While human dexterity has multiplied manifold with the advent of Generative AI, the world has also been plagued by a digital pandemic known as deepfake.
The threat is real, and AI experts are urging everyone leveraging the emerging tech, from intelligence and law enforcement agencies to banks and other institutions, to be a step ahead in keeping malicious attackers at bay.
Despite this, there isn’t enough awareness around deepfake, which raises a key question — Are we doing enough to nip this digital demon in the bud? Well, not exactly.
A recent PR stunt by the President of France, Emmanuel Macron, where he used AI-generated deepfake videos to publicise the commencement of the AI Action Summit in Paris, rather demonstrates that we have been trying to normalise the use of deepfakes, eventually blurring the lines between what’s real and what’s not.
Back home too, the deepfake menace is growing without bounds, unabated, and unchecked. Deepfake videos of eminent businessmen Mukesh Ambani or Narayana Murthy promoting various fraudulent trading platforms could be easily found doing rounds on social media platforms and WhatsApp groups.
Amid this, a recent report by Pune-based deepfake detection technology provider, pi-labs, underlines that the country’s BFSI sector, in particular, is exposed to the growing threat of deepfake.
Anticipating dangers that could be lurking ahead, the 2023-founded GenAI startup is endeavouring to help the global BFSI sector see through the smoke and mirrors by unmasking deception with its deepfake detection technology stack.
Founded by Ankush Tiwari, Abhijeet Zilpelwar, and Raghu Sesha Iyengar, pi-labs is building autonomous AI agents for deepfake detection, video analytics, and cybersecurity to save the BFSI sector from deepfake attacks.
But why BFSI?
Well, CEO and founder Tiwari feels that there is a significant lack of awareness about deepfakes among banks and enterprises.
Besides, the company’s report shows that before 2000, cyber frauds in India were largely restricted to prank videos and celebrity spoofs. By 2003, incidents of political manipulation and influencer-targeted pornography started coming to light.
Lately, cybercrime has largely become about financial frauds and scams around fake video KYC and digital arrest. From identity theft, loan fraud, money laundering, and manipulation of digital banks to fraudulent investment scams, the BFSI sector faces a tremendous amount of risk and potential losses due to deepfake.
It projected that India is on the verge of losing INR 70,000 Cr in 2025 alone to deepfake frauds.
At a time when there is a growing emphasis on Sovereign AI for India to be fully reliant on its own tech stack across the AI value chain, pi-labs has intensified its fight against deepfake.
Behind pi-labs’ Inception Story
All three founders of pi-labs carry decades of deeptech experience under their belt. Tiwari’s previous venture, Mobiliya Technologies, specialised in building mobile devices for defence use cases. Here, Zilpelwar was the VP of technology and Iyengar engineering director.
When Tiwari exited Mobiliya in 2019, after its acquisition, he knew that the next big technology opportunity was in the AI space.
As a tech entrepreneur, who has been developing tech for defence use cases, he decided to build a defensive stack to be ready from the early days as AI became mainstream.
“We have a simple philosophy. As new exponential tech emerges – which used to come every seven to eight years and now comes every five years, and going forward this time frame might get shorter – independent of the tech, cyber security is a constant. When quantum computing becomes big, cyber security needs are still going to be there,” said Tiwari, highlighting his thought behind incorporating pi-labs.
After its incorporation in 2023, the startup remained in stealth mode for a year. It came out in 2024 with its deepfake detection products — Athentify, pi-scout and pi-sense.
The startup claims to be the largest deployer of deepfake detection technology in the country. According to the CEO, the country’s BFSI space is too vulnerable to cyberattacks. Therefore, they are currently deploying their products with central and state law enforcement agencies in the country, including defence organisations. Currently, it serves eight such bodies.
According to the founder, the startup’s tech stack can detect manipulated content in any format—video, audio, images, and text. Besides BFSI, the bootstrapped startup is also looking to cater to HR use cases.
pi-labs’ Tech Stack
The startup first built Athentify, which it calls an “AI++-powered platform” that can detect and analyse deepfakes in different media formats. It uses multiple advanced AI algorithms to help its users identify synthetic content and fraud attempts using deepfakes.
This AI-powered platform is capable of analysing various signs of manipulations, including subtle inconsistencies in lip synchronisation, facial movements, voice patterns, voice clones, and more, hence raising alarms on any attempts of deepfake frauds.
“In short, we are using AI to beat AI, that’s our theme,” said Tiwari.
Besides, the startup’s other product, pi-scout, is a forensic AI agent that helps intelligence agencies get more detailed reports from various kinds of structured and unstructured data.
Explaining the tech, Tiwari said, “If you see any criminal investigation programmes today, most of these start with mobile phone data – call records, location data, pictures, CCTV footage, and bank statements. All these are various kinds of data. Now, how do you put all these together and make sense of them? So, you can think of pi-scout as a forensic AI agent.”
pi-sense does almost the same thing but decoding videos. It’s a video forensic analytics platform.
“In India, every forensic lab is sitting on a pendency of a few years, just because there are not enough people (who can work on them). And imagine that if a CCTV video analysis takes one hour, these forensic labs get videos of thousands of hours in a week in a lab. This is where we are using AI to help remove pendency,” said Tiwari.
Meanwhile, pi-labs claims to stand out from the competition with the use of India-specific data. Currently, its products support 15 Indian languages.
“If you look at any AI agent and ask whether it is a drug or not, it will not be able to be identified. If there is a picture of a murder, most LLMs will refrain from responding to that, but at pi-labs our job is to make sure our LLM plays that. Our problem starts from there,” Tiwari said.
Giving an interesting example, he said that if there is a drug investigation in Punjabi, AI must know the word Chitta, a synthetic drug made from heroin. No LLMs today can crack the word, but the way pi-labs has customised the LLMs to build its products, it can detect such words.
The Path Ahead For pi-labs
Largely catering to law enforcement agencies in the country, pi-labs is now also looking to onboard a few enterprise customers this year. International expansion is also in its playbook in 2025.
Currently, the startup is in talks with a few companies in the BFSI sector in North America.
Overall, the startup expects to grow its customer base to 20 by the end of the year. With this, it also expects to see a 3X jump in revenue from FY24 to FY25. However, the company did not reveal its revenue figures.
As AI evolves rapidly, cybersecurity experts, institutions, and governments face mounting challenges. The global deepfake AI market is projected to grow at a CAGR of 44.5%, surging from $564 Mn in 2024 to $5,134 Mn by 2030.
The Indian government’s INR 10,372 Cr IndiaAI Mission stresses the need to ensure ethical AI. As part of it, MeitY has sought proposals from startups and other relevant stakeholders for building AI-powered tools to detect deepfakes.
But to fight deepfake at its root, every institution has to become more vigilant, there is a need for more native AI companies to build deepfake detection tools. Though efforts have begun, India still lacks the necessary infrastructure.
This is where companies like pi-labs have a crucial role to play. With a strong hand-holding with its competitors like Kroop AI and Europe-based Sentinel AI, pi-labs is sitting on a mine of opportunities going ahead. For now, can pi-labs help the BFSI sector fend off the deepfake menace with its AI arsenal?
[Edited By Shishir Parasher]
The post Can pi-labs Fend Off Deepfake Attacks With Its AI Arsenal? appeared first on Inc42 Media.