Day 3 - AI Impact Summit

Day 3 of the AI Summit generated very interesting conversations yet, all rooted in the core principles which have highlighted in Day 1 and Day 2 of the AI Summit i.e. accountability, trust, and governance. While the sessions spanned a wide spectrum of themes, what stood out most for us as lawyers were the discussions on the future of fair tech, emerging governance and fairness models, and the evolving landscape of AI and intellectual property. These sessions not only sparked debate but also offered valuable insights into how legal, ethical, and technological frameworks must evolve together in an era of rapidly advancing AI.

Safe & Trust AI : the ethics & governance perspective

AI is increasingly used across industries, making it essential for stakeholders to define boundaries for its responsible deployment. Since AI systems depend entirely on the data fed into them, high‑quality and accurate data is crucial—especially in public‑impact sectors—to prevent errors and hallucinations. India must prioritise identifying critical sectors and deploying AI carefully within them.

Stakeholders also carry a duty of care to ensure AI does not cause harm. They must spread awareness, as lack of understanding often leads to ethical lapses. In the legal world, AI hallucination has become a serious issue. Recently, the Supreme Court of India warned against the improper use of AI for drafting pleadings and legal research after encountering fabricated case laws and non‑existent judgments in filings—actions the Court said erode trust in judicial processes. Also, the data on which the AI response are trained must be latest and authentic to generate trust. Users must therefore verify AI‑generated information independently.

Because AI models inherently reflect biases in their training data, users must check outputs for accuracy, bias, and reliability, and disclose whenever AI is used. Continuous data updates are essential, as outdated information limits AI’s accuracy.

From a governance standpoint, the state must embed strong ethical frameworks. In healthcare, for example, ICMR’s Ethics Guidelines and government policies on privacy, accountability, and liability support responsible AI integration in hospitals.

Also, discussion veered towards agentic AI systems which create distributed responsibility, and not absolution of responsibility. Responsibility cannot be delegated to the AI itself, and “the agent did it” is not an acceptable gap. Accountability has to be explicitly layered across the AI lifecycle from model developers and system integrators to organizational decision-makers and deployers. Preventing accountability “evaporation” was highlighted as a central governance challenge for agentic systems.

While global principles (transparency, accountability, proportionality, rule of law) are broadly aligned across jurisdictions, the panel emphasized that principles without tooling are ineffective. There was strong emphasis on lifecycle-based AI governance (pre-deployment to post-deployment), practical audits, AI impact assessments, and standards-based management systems and continuous monitoring for misuse, and emergent behaviours in agentic systems.

Future of Fair Tech : Addressing Equity, Safety and Accountability in a rapidly evolving AI landscape

The discourse revolved around the gap between high‑level principles of fair and responsible AI and the difficulty of converting these values into practical systems, standards, and real‑world outcomes. Although these principles are presented as universal, implementation remains inconsistent and uneven across contexts.

Three major global tensions were highlighted: unequal access to AI capabilities, fragmented accountability structures, and declining public trust. Fairness was framed not as an abstract value but as a technical and operational requirement tied to accessibility, safety, and the need for more representative datasets.

Governance emerged as the core issue. Effective governance must be embraced collectively by industry, governments, and civil society. In the media economy, AI‑generated summaries and zero‑click consumption are eroding the financial stability of news organizations, shifting advertising revenue—and therefore influence—toward major digital platforms. Stronger collaboration between AI developers and media outlets, including fair compensation for content used to train AI systems, was emphasized.

Since information underpins democracy, AI‑driven filtering risks accelerating information decay, removing context, and concentrating narrative power in a few global platforms—raising concerns of digital colonization. Finally, while AI functions through global supply chains, governance remains fragmented. A one‑size‑fits‑all regulatory model is impractical; responses must be locally grounded yet supported by shared global guardrails to effectively manage diverse risks.

Charting India’s AI-IP Playbook

India can position itself as a leader in shaping the AI–IP landscape by regulating AI effectively and building a strong IPR framework that fosters trust and innovation. It was noted that AI can significantly improve the grant of IPRs, especially in patents, where extensive manual searches for prior art could be handled faster, more accurately, and at lower cost through AI‑driven tools. Since IPR fundamentally rests on data, technology, innovation, and trust, AI systems must be designed to support—not replace—human judgment while ensuring innovation is protected, bias and discrimination are avoided, and privacy remains intact.

The discussions emphasized the need to balance the beneficial and harmful effects of AI, recognizing that efficacy of existing laws such as the IT Act, 2000 and the DPDP Act, 2023 need to be looked into. Any future AI law should be developed gradually rather than hastily. Creating an environment that accelerates and democratizes innovation through openness and transparency was highlighted as essential, along with embedding IPR protections that secure downstream creation. The DPIIT’s proposal for a mandatory blanket license—requiring AI developers to pay royalties into a central fund—sparked debate, with NASSCOM opposing it and advocating instead for a text‑and‑data‑mining exception similar to Japan, Singapore, and the EU. This led to broader discussions on IP challenges in AI training and outputs, including unresolved questions around fair use versus fair dealing, authorship, ownership, liability, and corporate authorship, as seen in the ongoing ANI vs. OpenAI case. Personality rights also remain legally unclear. It was concluded that without legislative amendments allowing text and data mining, AI development will remain constrained, and creators should always have the choice to opt out of having their work used for training. Ultimately, meaningful dialogue is needed to strike the right balance between promoting innovation and protecting IPRs.

On 3^rd day of AI Impact Summit, Sarvada Legal attended:

1.    Safe & Trust AI : the ethics & governance perspective

2.   Sovereign AI Infrastructure for Bharat and Global South

3.   Future of Fair Tech : Addressing Equity, Safety and Accountability in a rapidly evolving AI landscape

4.   Safe and trusted Agentic AI : Building accountability and inclusion for India and the Global South

5.   Charting India’s AI-IP Playbook : Innovation, Rights and National Advantage

6.   Data for Development : Building AI in the Global South.

Please feel free to reach out to our Team to discuss any of the Technology Law, Competition Law, International Trade and Policy Issues.