Two of the Artificial Intelligence projects under the India AI Mission will be ready to showcase to the government by 2026, said startups involved in the process.
One of these projects, Jivi AI is already available to the public since October 2024, but will be integrated into public hospitals as per the proposal submitted by the company for the Mission. Another India AI project expected to be ready in 2026 is ParakhAI, one of the eight Responsible AI-themed projects under the Mission. Proposed by Civic Data Lab research organisation, ParakhAI seeks to provide a public tool for citizens to flag any unethical work done by an algorithm or AI.
Health Assistant
The Jivi AI app is a voice-first doctor, created by Jivi Health Pvt. Ltd, that acts like a co-pilot for a real clinician or personal AI health assistant to consumers. According to Ankur Jain, Founder of Jivi, healthcare faces three main problems: availability of doctors, quality of doctors and cost to get healthcare. Jivi provides help in this regard, particularly rural government hospitals and primary healthcare centres where most workers are junior nurses or doctors.
“We can give AI applications to the workforce based on government protocols. This increases the quality of work, improves the health outcome of patients all over India. That was our thesis when we approached the government. They really liked our vision,” said Jain. Once the Jivi app is licensed by the hospitals, it will be integrated with the hospital’s electronic health record (EHR) systems. The government has offered the startup around ₹25 lakhs to the start-up to build such a prototype, said Jain.
The AI model will also be incorporated into the Ayushman Bharat Digital Mission (ABDM), a digital health infrastructure initiative of the government. This will allow people to upload their health reports to and from their ABDM account when using the Jivi AI.
“We can fetch the information from there as well. People can upload their health reports in an ABDM locker and anybody can download the same after getting user consent,” said Jain.
When asked about regulatory hurdles, Jain said the app is not recognised as a medical device and thus does not face as many regulatory challenges. However, keeping in mind privacy laws in India as well as biases within datasets, the company also created its own synthetic data for the AI. It may be noted that the data collected by the AI is still stored in Jivi databases.
Jain said the company will have a workable version ready for hospitals by the first or second week of May 2025. This will be followed by another set of evaluation, following which it will be integrated by the ministries.
Auditing AI
“When it comes to high risk AI application, where a citizen’s rights get impacted, it becomes critical to audit the algorithms taking these decisions in an automated manner. This could be in terms of employment, food, public health, insurance. That’s where we need systems like ParakhAI, where citizens can audit the algorithms and report any biases or exclusions,” said Gaurav Godhwani, Executive Director and Co-Founder of Civic Data Lab.
He argued that the idea of algorithmic auditing should be thought from the development stage to deployment and testing stages rather than as a tail-end process.
With the project still in the developmental stage, Godhwani said the team is still deliberating on whether to provide the tool as a chatbot or as another application.
“It could be a checklist or a chatbot where citizens can also give their inputs on a timely basis and have a much better reporting mechanism for explaining the issue to AI developers and regulators,” said Godhwani.
When asked about the timeline, Godhwani said the project will likely be complete by mid-2026. Most of the projects of safe and trusted pillars are longer in nature because of a lot of iterations, discussions, said the company leader. For now, Civic Data Lab is starting a focus working group, working with stakeholders and experts from the AI community to shape the framework. Further, Godhwani said that the ParakhAI, once-ready, will be among the first government-funded responsible AI framework that allows participatory AI auditing work and learning from the community.