Sitharaman calls for stronger AI safeguards without stifling innovation

Mumbai: Finance Minister Nirmala Sitharaman on Friday called for stronger safeguards around artificial intelligence as the use of agentic AI expands, while cautioning regulators against stifling innovation.

Speaking at the Global Fintech Fest 2026, Sitharaman said that the rapid rise of agentic AI is creating risks that cannot be left for society to deal with after the technology has moved ahead.

“I raise this not to cause alarm, but to emphasize institutional accountability towards larger humanity,” she said, questioning whether the AI industry was doing enough to reassure the public about the safety of increasingly powerful systems.

“Who from the global AI industry will stand up and reassure the public?” she asked. “Who will demonstrate that safeguards are actually keeping pace with the speed? Where is the structural collective effort to provide credible and transparent answers?”

She said technology itself could engineer solutions, but warned that safeguards needed to evolve quickly and continuously. “Unless those safeguards arrive quickly and update continuously, the question of AI going rogue is left for the society at large to answer. That is an unsustainable distribution of risk,” she said.

Sitharaman said she had discussed with the Reserve Bank of India whether a regulatory approach could address emerging AI risks without hurting innovation.

She stressed that the challenge was particularly important as the financial system moves into an era of agentic AI, which can move beyond recommendations to taking direct actions.

“AI acts as a double-edged sword. It helps us spot fraudsters and fraud faster, but it also allows attackers to automate larger attacks with more sophisticated cyber-attacks,” she said.

The finance minister called for a regulatory approach that distinguishes between unnecessary friction and safeguards that are essential to the system.

“Our goal should be neither to block innovation out of fear nor to let speed compromise safety,” she said. “Instead, we must separate wasteful operational friction from essential safeguards, the built-in checks, human oversight, and circuit breakers that protect the system.”

She said the responsibility for AI could not be delegated entirely to technology teams in a company or a bank. “AI may assist judgment, but responsibility must remain human and institutional,” she said, adding that the more serious the consequence of a decision, the stronger the requirement for review, explanation, and opinion.

Institutions must understand where AI creates value, where it creates material risk, and what safeguards are required in each use case, she said

Higher-risk applications should face greater scrutiny before deployment and throughout their lifecycle, while the broader policy objective should be to build systems that are fast yet accountable, autonomous yet reversible when mistakes occur, and innovative without diluting consumer trust. “Efficiency and safety should not be competing objectives,” she said. “Good institutional design must make them mutually reinforcing.”