Ankush Sabharwal

#ExpertSpeak | #ORFIndia 👩‍💻 The Right to Be Forgotten faces fundamental challenges in #AI systems that internalise data rather than store it. #India’s DPDP Act must reinterpret erasure as functional non-influence, not literal deletion. Embracing pragmatic machine unlearning can balance privacy rights with technical feasibility and sustainable #AI innovation. Tanusha Tyagi argues that #India’s Right to Be Forgotten must evolve beyond data deletion to address how personal data is internalised, retained, and inferred by #AI systems ➡️ https://or-f.org/37054

9

SANATAN LABS - Reputation Risk Audit of “Indigenous AI” (PASSIONIT PRUTL KALKI AIDHARMA framework) An Audit by Ganesha, with Krishna and Kalki In a quiet digital mandap, Ganesha — remover of obstacles and lord of intellect — opens a scroll. Krishna smiles. Kalki activates a transparent dashboard. The topic: “Indigenous AI” — Claim vs Capability. 1 Definition Risk: What Does “Indigenous” Mean? Ganesha begins: “Before building, define.” There are four layers of AI sovereignty: Application Layer – Using external APIs to build local apps Fine-Tuned Layer – Customizing open-source models Foundational Layer – Training large models from scratch Infrastructure Layer – Owning chips, compute clusters, data pipelines Risk arises when: Layer 1 or 2 is marketed as Layer 3 or 4. Terminology blurs to create perception advantage. 2 Reputation Risk: When clarity emerges globally, credibility erodes. Kalki observes: “If compute scale is limited, but ambition is unlimited, narrative may outrun engineering.” Signals of capability mismatch: No published model benchmarks No peer-reviewed papers No open evaluation datasets No transparent parameter disclosures Krishna adds: “Confidence without validation becomes noise.” 3 Global Benchmark Risk Indigenous AI will be measured against: Multilingual performance Safety alignment Latency and inference cost Hallucination rates Developer ecosystem adoption If global users test and find gaps: Reputation damage spreads quickly. AI credibility compounds slowly — collapses rapidly. 4 Governance & Data Risk Ganesha lifts another scroll: Questions global observers will ask: Who owns the training data? Is consent documented? Are datasets audited for bias? Are model weights accessible or opaque? If “sovereign AI” still depends on foreign cloud infra, hidden dependencies weaken the claim. 5 Narrative Amplification Risk Krishna points gently: “When announcements exceed execution milestones, video narratives outpace reality.” Over-celebration creates: Inflated expectations Investor pressure Public skepticism Talent attrition if hype collapses 6 Policy Alignment Risk True indigenous AI must align with: National data protection frameworks Ethical AI standards Long-term semiconductor strategy Academic research ecosystem If policy support is symbolic rather than structural, sustainability weakens. 7 The Trust Equation Ganesha summarizes: Reputation = Transparency + Technical Depth + Measurable Impact – Exaggeration Without transparency, speculation grows. Without measurable impact, claims decay. Krishna’s Closing Insight “Indigenous AI is not a slogan. It is an ecosystem.” He continues: Build quietly. Publish rigorously. Benchmark honestly. Scale responsibly. Kalki’s Final Warning “The future will not punish ambition. It will punish misalignment between narrative and reality.” Ganesha closes the audit: “Remove obstacles — but first remove ambiguity.” Because in AI, Trust is the real sovereign asset.

1

"The Office of the Principal Scientific Adviser (OPSA) to the Government of India has released a White Paper titled “Strengthening AI Governance Through Techno-Legal Framework”, outlining India’s approach to building a trusted, accountable, and innovation-aligned artificial intelligence (AI) ecosystem. The White Paper emphasises the contours of “techno-legal” AI governance framework that aims to mitigate risks while preserving flexibility and innovation. It highlights India’s pro-innovation approach to AI governance, which integrates baseline legal safeguards, sector-specific regulations, technical controls, and institutional mechanisms. “Developing a robust and responsive governance framework is not just a regulatory necessity but a prerequisite for sustaining the momentum of technological progress. The techno-legal approach offers a viable pathway by embedding legal, technical, and institutional safeguards into AI systems by design,” said PSA Prof. Ajay Kumar Sood during the release of the white paper. The White Paper defines the techno-legal approach as a practical and ecosystem-wide model that embeds governance directly into the design and operation of AI systems by default. Key focus areas covered in the White Paper include understanding the techno-legal approach to AI governance; enabling safe and trusted AI across the full AI lifecycle; technological pathways for operationalising techno-legal governance; implementation considerations for India’s AI governance framework; and the development of techno-legal tools and compliance mechanisms. This publication is the second in the White Paper Series on ‘Emerging Policy Priorities for India’s AI Ecosystem, an initiative by OPSA that aims at deepening understanding and fostering informed discussion on critical AI policy issues. The first White Paper in the series, released in December 2025, focused on “Democratising Access to AI Infrastructure”, highlighting the need to treat AI infrastructure as a shared national resource and identifying key enablers such as access to high-quality datasets, affordable computing resources, and integration with Digital Public Infrastructure (DPI). These White Papers, framed as explanatory knowledge documents, are intended to support informed deliberations on various policy priorities to shape the evolving AI ecosystem, and to reinforce India’s catalytic role in shaping the global AI governance discourse."

93

6 Comments

#DateExtended | Apply by 25 March | #Callforproposals IndiaAI, in partnership with the Real Time Governance Society and Line Departments of the Government of Andhra Pradesh, invites innovators to co-create AI-driven solutions that strengthen governance and enable inclusive growth. The opportunity is focused on 6 Strategic Domains: #Empoweringruralentrepreneurs #Optimisingsupplychains for essential public services #Streamliningrenewableenergy land allocation #Urbaninfrastructure planning #Landusemonitoring and compliance #Predictiveinfrastructuremanagement for government schools This initiative aims to deliver measurable outcomes including skilled professionals, research-led innovation, real-world pilots, and stronger collaboration between government, academia, industry, and startups. Key Dates Launch: 23 January 2026 Submission Deadline: 25 March 2026 Final Cut-off: 31 May 2026 Eligible #AIStartups, #researchers, and #innovators are invited to submit #proposals via the #ManthanPlatform. Application Link: https://lnkd.in/g_HjmU6g Please note: Applicants must register as #SupplySide on the Manthan platform to apply. We look forward to building #AI that transforms #governance and delivers real impact—in collaboration with the #GovernmentofAndhraPradesh. Office of the Principal Scientific Adviser to the Government of India | Manthan O/o PSA | Dr Sapna Poti | KAVITA T. | Nida Faruqui | Srinivas Kancharla IIMc, PMP | Tushar Jha | Sai Sankar Sarangi | Githin Chandran | Sonu Kumari #IndiaAI #TransformingGovernance #GovTechIndia #ArtificialIntelligence #ML #NLP #AIAgents #DigitalGovernance #DigitalIndia #RuralInnovation #InnovationAtScale #MSME #AndhraPradesh

12

Unified Agent Protocol (UAP) to Enable AI Agents in UPI for Small, Frequent Purchases Overview A new framework named the Unified Agent Protocol (UAP) is being developed to support AI agents operating within the UPI ecosystem. The aim is to enable automated interactions and payments through AI agents, with an initial emphasis on low-value, frequent purchases. Initial use case In its first phase, UAP is expected to target routine transactions such as groceries, where small, frequent payments can be processed more efficiently. More sophisticated use cases could emerge later as technology matures. Potential implications For regulators and policymakers, the deployment of AI-driven payment agents within the UPI framework may necessitate attention to data privacy, security controls, and consumer protections. For businesses and consumers, the framework could reduce friction in small purchases and streamline interactions within the UPI ecosystem. For the investment community, the development of UAP may influence interest in AI-enabled payment solutions and related technology platforms.

1

"The India AI Governance Guidelines" Released in February 2026, it establishes a principle-based framework designed to foster safe, trusted, and inclusive AI innovation across the country. This framework adopts a "whole-of-government" model that prioritizes innovation over restraint, positioning artificial intelligence as a primary catalyst for achieving the vision of "Viksit Bharat 2047". The Seven Guiding Sutras: The framework is anchored in seven core principles, or "sutras," which serve as the foundation for responsible AI development: 🔑 Trust is the Foundation: Trust must be embedded across the entire value chain, from the technology itself to the institutions supervising it. 🔑 People First: AI systems must strengthen human agency and maintain meaningful human oversight. 🔑 Innovation over Restraint: Responsible innovation is prioritized over cautionary restraint to achieve national socio-economic goals. 🔑 Fairness and Equity: Systems should be designed to avoid bias and discrimination, particularly against marginalized communities. 🔑 Accountability: Responsibility is assigned based on the function performed and the risk of harm. 🔑 Understandable by Design: AI systems must have clear explanations and disclosures to help users understand how they work. 🔑 Safety, Resilience, and Sustainability: Safeguards should minimize harm, and development should be environmentally responsible and resource-efficient. ✅ The Six Pillars of Recommendations The guidelines examine key issues and offer recommendations across six strategic pillars: 1. Infrastructure: Leveraging the IndiaAI Mission to provide subsidized access to compute (aiming for 100,000 GPUs) and the AIKosh platform for datasets and sectoral models. 2. Capacity Building: Scaling initiatives like IndiaAI FutureSkills to train a future-ready workforce and expanding AI literacy to grassroots levels through AI Data Labs. 3. Policy & Regulation: Adopting an agile approach that builds on existing laws (like the DPDP Act, 2023, and IT Rules) while identifying gaps for targeted legislative amendments regarding liability and content authentication. 4. Risk Mitigation: Developing an India-specific AI risk assessment framework and a national, federated AI incident reporting mechanism. 5. Accountability: Implementing graded obligations and liability proportional to the risk and function of the AI actor, alongside mandated grievance redressal mechanisms. 6. Institutions: Institutionalizing a "whole-of-government" model through new bodies:    ✅ AI Governance Group (AIGG): To coordinate overall policy development.    ✅ Technology & Policy Expert Committee (TPEC): To provide expert advisory.    ✅ AI Safety Institute: To conduct research, develop standards, and provide technical guidance. Source: https://lnkd.in/gDzJ2RGA

2

The India AI Impact Summit signals a structural shift. AI governance is moving from principles to infrastructure. Look at the agenda: • Sovereign and shared AI compute models • Global South safety and evaluation networks • Multilingual AI safety benchmarks • Incident monitoring and early warning systems • Agentic AI governance stacks • Cross-border standards and codes of practice • AI quality infrastructure for trade and certification This isn’t another ethics conversation. It’s about building the enforcement and assurance layer of the AI economy. India has defined its AI Governance Guidelines and Seven Sutras. The real test now is execution. Can trust be engineered into production systems? Can governance run with AI — not after it? Because in the next phase of AI leadership, governance capacity will matter as much as model capability. We’ll be engaging in this dialogue in New Delhi (Feb 16–20). https://lnkd.in/gV4nsur2 #AIGovernance #ResponsibleAI #GlobalSouth CognitiveView

8

Kuba Szarmach you may want to look a little closer rather than the emdash filled LLM AI slop that you posted. An honest human assessment exposes a very different perspective. India’s January 2026 white paper on “Strengthening AI Governance Through a Techno-Legal Framework” deserves more respect than most government AI documents. It is not empty hype. It shows real work, consultation, and intent to move beyond ethics theatre into technical enforcement. Rightly, the paper rejects the fantasy that AI can be governed by law alone. It pushes governance into system architecture across the lifecycle: data, training, inference, and agentic systems. It treats agentic AI as a distinct risk class requiring identity controls, monitoring, and kill switches. That is rare and important. It also recognises that deepfakes cannot be solved by takedowns alone, instead targeting the pipeline through provenance, persistent identifiers, and infrastructure-level obligations. It anchors governance in India’s Digital Public Infrastructure: Aadhaar, DEPA, DigiLocker, UPI, consent rails, and clean rooms. This attempts to shift control from vendor policy to state-level infrastructure. The technical stack referenced is substantive: machine unlearning, privacy-enhancing technologies, synthetic data, bias audits, red teaming, confidential computing, provenance tooling, and runtime guardrails. Now the hard truth. This is governance modernisation, not sovereignty. There is no hard guarantee that India can survive vendor withdrawal, foundation model cutoff, compute denial, API lock-in, or cloud jurisdiction overrides. Lifecycle controls are discussed while supply chain dependence and hyperscaler leverage remain largely untouched. It leans heavily on voluntary commitments, soft standards, and industry goodwill. That produces compliance optics, not enforceable authority. Risk controls are described, but rarely mandated. Audits, red teaming, monitoring, and incident reporting are proposed without clear legal teeth, penalty triggers, or non-negotiable runtime veto power. It builds legibility before building command. We get taxonomies, committees, safety institutes, and incident databases. What we do not get is runtime supremacy, update authority, or guaranteed withdrawal survivability. Cross-border risk is acknowledged, but there is no credible mechanism to stop foreign policy or foreign law from overriding domestic intent. This paper prepares the narrative for control without yet securing control. It is permission choreography more than authority. Valuable raw material, yes. But it does not yet break dependency or make AI infrastructure resilient under geopolitical stress. If India wants real AI sovereignty, the next step is not more frameworks. It is enforceable control planes, mandatory kill tests, compute and model independence, and legal authority that operates at runtime rather than in PDFs. #AIGovernance #DigitalSovereignty #ResponsibleAI #AIInfrastructure

26

1 Comment