As artificial intelligence becomes embedded deeper into enterprise decision-making, responsible AI cannot be treated as a standalone technology or compliance exercise. Organisations need governance models that account for how AI is deployed, the consequences of automated decisions, the jurisdictions in which systems operate and, crucially, who remains accountable when something goes wrong.
That was the major theme of the panel “Responsible AI & Governance: Ethics, Bias, Compliance and the Global Regulatory Landscape” at the ETLegalWorld AI-Powered Legal Transformation Summit 2026. Moderated by Jyoti Pawar, Independent Counsel – AI, Data & Tech, it featured Saurabh Awasthi, VP & Sr. Counsel, India, Kyndryl Solutions; Jaswinder Singh Shekhawat, Head Legal & Compliance, MediBuddy; Sudha Hooda, Executive Director, Legal, NVIDIA; and Deepalakshmi Vadivelan, General Counsel & SVP Legal, Global DPO, Quess Corp.
Pawar framed the discussion around an expanding list of governance questions including autonomous decision-making, model opacity, cross-border regulatory conflicts, data provenance, and who bears responsibility for an AI-generated outcome.
Boards need AI literacy, not just policies
Sudha Hooda, speaking in her personal capacity, said enterprises should first prioritise improving AI understanding at the board level rather than creating a standalone AI committee. “I do not believe that there needs to be a separate committee which works in a silo,” she said, arguing existing board committees should build enough AI capability to assess impact within their own remit.The bigger risk, she said, is insufficient understanding of the technology itself. She urged distinguishing repetitive tasks that can be automated from consequential decisions still needing human judgment, and separating regulation of computing capability from regulation of its applications, comparing AI to electricity, where governance should target how power is used, not the infrastructure itself.
For Jaswinder Singh Shekhawat, risks are sharpest in healthcare and insurance, where AI aids underwriting and claims but also touches sensitive data and consequential decisions. He flagged bias from historical training data reproducing demographic skews, making human oversight essential, and data privacy around storage and access as key concerns.
The deeper question, he said, is accountability – organisations must pre-decide whether responsibility for AI-caused harm sits with management, product, legal or infosec, rather than resolving it after an incident.Deepalakshmi Vadivelan said enterprises must first understand their position in the AI value chain. Developers may justify a dedicated governance committee given heightened design responsibilities; deployers can fold AI risk into existing risk, audit, cybersecurity or ethics structures. This distinction matters because enterprises can drift from deployer to provider status as they customise AI with their own data, altering their legal obligations.
Responsible AI: not a checklist
Responsible AI cannot be a checklist owned solely by legal or compliance, the panel concluded. Boards need AI literacy for real oversight; legal teams need technical fluency; business functions need clarity on when automation can be trusted and when human judgment must intervene. The right framework will vary by industry, use case, geography and an organisation’s place in the AI value chain.
For Pawar and the panelists, the challenge isn’t finding the right regulation, it’s building governance that keeps pace with technology while keeping accountability identifiable, human oversight meaningful, and innovation from outgrowing the systems meant to keep it trustworthy.
The AI-Powered Legal Transformation Summit 2026 by ETLegalWorld is currently being held in Bengaluru. The summit has brought together general counsel, in-house legal leaders, and technology innovators, to chart the next phase of legal transformation in an AI-driven world.
