AI regulation in India has moved from a distant policy debate to an urgent national question. Deepfakes, AI-assisted cyberattacks and biased hiring tools are testing whether existing safeguards can keep up. Reports suggest the government has begun preparing a consultation paper on regulating artificial intelligence.
In a recent News18 explainer on whether governments can control AI before it is too late, Jaspreet Bindra, co-founder of AI&Beyond, made one of the sharpest arguments in the debate. The companies building powerful AI, he argues, should not be the only ones deciding how risky it is.
His position rests on four ideas: self-reported risk is a conflict of interest, testing must be truly independent, rules should scale with risk, and India must build its own capacity to check AI systems. Here is what each one means, and why it matters for the future of AI regulation in India.
Why AI regulation in India is on the agenda now
Pressure for stronger oversight has grown on several fronts. AI companies themselves have issued warnings about the risks of their technology. Some companies reportedly expect catastrophic cyberattacks within the next six to 12 months.
For India, the challenge is a balance. The country wants safeguards that protect citizens and address new risks, without stifling the research, investment and innovation that make AI valuable. Bindra’s proposals are aimed squarely at that balance.
1. Self-reported AI risk is a conflict of interest
For years, AI safety has largely run on trust. Companies run internal evaluations, publish reports and decide what to share with outside researchers. Bindra says that letting AI companies choose which risks to disclose creates “a clear conflict of interest.”
His concern is selective disclosure. A business may highlight the risks it can easily manage while playing down the serious ones. Bindra names three examples: unauthorised agent actions, data leaks and algorithmic bias.
The first example is telling. As agentic AI moves from answering questions to taking actions, an agent doing something nobody approved becomes a real business risk. Bindra also points out that voluntary disclosures vary from company to company, which makes fair comparison almost impossible for regulators and buyers alike.
2. Independent AI testing must be genuinely independent
Bindra supports independent safety testing, but with conditions attached. In his view, testing only counts if it works to clear standards and produces results that others can verify. Companies, he argues, should not have complete control over how their own risks are assessed or explained to the public.
The explainer makes a related point. Access granted at a company’s discretion is not real independence. Evaluators need the freedom to investigate risks and report serious findings without interference, while genuinely sensitive information stays protected.
This is the difference between a safety audit and a press release. Without mandatory reporting, a developer could find that its model helps with sophisticated cyberattacks, and regulators might never hear about it.
3. A risk-based AI framework, not one rulebook for every model
Bindra’s central recommendation is that India should regulate AI according to the risk it creates. Applying identical rules to every model would either over-burden simple tools or under-protect dangerous ones.
Under his approach, low-risk applications could run under lighter guidelines and regulatory sandboxes. AI used in healthcare, finance, employment and law enforcement would face stricter requirements for transparency and human oversight.
The logic is easy to see. A chatbot that answers store-timing questions and a model that approves loans or screens job candidates carry very different stakes. A risk-based framework also protects startups: the explainer warns that expensive audits and repeated approvals tend to favour large technology companies that can absorb the cost.
4. Build India’s own AI testing capability
Bindra’s fourth point turns from rules to capacity. India, he says, must build independent testing capabilities through the IndiaAI Mission, the IITs and IISc.
He pairs this with two tools: open benchmarks and third-party audits. Together, they would reduce India’s reliance on AI companies to assess their own systems, while still leaving room for startups to keep innovating.
This matters because a regulator is only as strong as its ability to check claims. Fellow expert Roshan Farhan, founder of MyriadAI, adds a local angle in the same explainer: testing should cover performance across Indian languages and demographic groups. Home-grown testing infrastructure is the most practical way to make that happen.
How Bindra’s view fits India’s AI governance framework
India is not starting from zero. The AI Governance Guidelines from the Ministry of Electronics and Information Technology (MeitY) rest on seven guiding principles, including trust, human oversight, fairness, accountability, transparency and safety.
Rather than a standalone AI law, the guidelines recommend using existing legislation, finding gaps and making targeted amendments. They also propose coordinated oversight through an AI Governance Group, a Technology and Policy Expert Committee and the IndiaAI Safety Institute.
The framework itself admits two weaknesses: voluntary arrangements lack legal force, and responsibility across developers, deployers and users is not clearly defined. Bindra’s proposals target exactly that gap, turning good principles into checks that can be tested and enforced.
What this means for businesses using AI in India
If AI regulation in India follows the risk-based path Bindra describes, companies deploying AI can prepare now:
- Sort AI use cases by risk. Anything touching health, credit, hiring or policing is likely to face the toughest rules.
- Keep a paper trail. The explainer notes regulators could require incident reporting within set deadlines and preserved records.
- Keep humans in high-stakes loops. Human oversight is a recurring theme in both Bindra’s view and MeitY’s principles.
- Ask vendors for proof. Prefer third-party test results over a supplier’s self-reported safety claims.
Frequently asked questions
What is Jaspreet Bindra’s view on AI regulation in India? He backs a risk-based framework: lighter rules and sandboxes for low-risk AI, stricter transparency and human oversight for high-risk sectors, and independent testing built through the IndiaAI Mission, IITs and IISc.
Why does Bindra say AI companies shouldn’t assess their own risks? Because companies may highlight manageable risks and downplay serious ones, such as unauthorised agent actions, data leaks and bias. Voluntary disclosures are also too inconsistent to compare fairly.
Does India have a dedicated AI law? Not yet. MeitY’s guidelines favour existing laws plus targeted amendments, and the government is reportedly preparing a consultation paper on AI regulation.
The bottom line
Bindra’s argument is not anti-innovation. It is about verification. Trust in AI, in his framing, should come from independent, standards-based testing rather than from the word of the companies selling it.
For policymakers, the message is to scale rules with risk and invest in testing capacity at home. Done well, AI regulation in India could protect citizens and still give Indian startups room to build.
Source: News18, “AI’s Evolving Capabilities Demand Stronger Regulatory Oversight”, Shilpy Bisht, 28 October 2026.