The rapid embracement of artificial intelligence (AI) in India is now revealing a dark underbelly. Companies are relying on illicit data to fuel their AI developments. They are sometimes acquiring data from unregulated dark marketplaces to deploy AI models faster than the rivals.
India has become a hotspot for data markets due to its vast population and data-rich environment. An anonymous startup employee revealed to one of the leading TV news media that the source of its training data is not being questioned. It is not asked as deadlines are crazy and the founders look for getting ready the models within weeks.
A similar experience was shared by a researcher from IIT, which is one of India’s top institutes. He resigned from a startup after learning how the data is sourced. The founder wanted an LLM in a few weeks and the pressure was at the expense of ethical standards.
Data is the lifeblood of AI and it enables the models to improve with more information they process. Feeding AI flawed or illegally sourced data results in unreliable systems which are riddled with biases. Umakant Soni of AIfoundry said that the biases of AI can scale exponentially and he warns that unchecked practices could lead to dystopian outcomes. He added that curriculums in schools are regulated to ensure sound education for humans and similar approach is required for AI too.
AI companies are facing scrutiny for using vast and unverified datasets across many countries. The challenge in India is more complex due to the linguistic diversity. Many AI systems lack data in regional languages and this force the startups to rely on unregulated sources. AI ethicist Jibu Elias said that the scarcity of machine-usable data in Indian languages reinforces biases as models still heavily depend on English.
Many datasets are also involving sensitive information like healthcare or financial records and this even bypass the privacy norms as well as damage public trust in AI technologies.
India risks stifling innovation and eroding public trust in AI without immediate action.
