Field notes on RBI model risk management, DPDP, SEBI, and building AI agents you can actually put in front of a regulator.
We have been demoing Sakshi over screen shares for months. Governance is not a slideshow, so we built sandboxes. Request your own Sakshi instance, seeded with syn...
RBI's draft Model Risk Management guidance is the most consequential AI regulation Indian finance has seen. We filed our formal comments this month. Here are the ...
India shipped agent-initiated UPI payments faster than almost anyone - and skipped the accountability layer that Google's AP2 protocol made its centerpiece. NPCI'...
Enterprises answered every AI risk question with 'there's a human in the loop.' That human now approves four hundred agent actions a day. The control exists on pa...
When an agent makes a bad decision, the post-mortem question is 'what was it thinking?' - and application logs cannot answer it. Reasoning capture has to be built...
Your bank can tell regulators everything about its customers and almost nothing about the AI agents already touching customer data. Agent identity - Know Your Age...
India just released its AI governance framework - principle-based, not prescriptive, and built on the RBI's FREE-AI playbook. Here's what the seven sutras mean in...
Three-hour takedowns. Mandatory AI labeling. Criminal penalties for synthetic media misuse. India just became the first major economy to regulate deepfakes with e...
A social network for AI agents just exposed critical vulnerabilities in multi-agent systems. Here's what it means for Indian enterprises - especially in regulated...
The emerging AI audit industry is largely checking boxes that don't matter. Here's what a real audit would look like - and why continuous monitoring beats point-i...
2025 was the year AI moved from pilots to production in Indian enterprises. Here's what worked, what didn't, and what we expect in 2026.
Nobody's doing the math on what happens when every enterprise app has AI embedded. The economics are brutal if you don't manage them proactively.
Retrieval-Augmented Generation was supposed to ground AI responses. Instead, it created new ways to fail. Here are the seven deadly failure modes we've catalogued...
While the West scrambles to build AI oversight from scratch, India already has the infrastructure. The layered regulatory framework everyone complains about might...
What the foundation model companies won't tell you about those impressive token counts. That 128K, 1M, or 2M context window? It's a theoretical maximum, not a pra...
Standard AI fairness tools check for race and gender bias. They miss caste, religion, region, and the intersections that matter in India. Here's a framework that ...
Your model worked great at launch. Six months later, performance has quietly degraded. Here's how to detect drift before it becomes a crisis.
Everyone talks about 'data staying in India.' But true AI sovereignty requires rethinking the entire stack - from model training to inference to governance.
Your model has 95% accuracy on the test set. Great. But accuracy doesn't tell you if it's reliable in production. Here's what to measure instead.
Your chatbot handles Hindi. It handles English. But when a user says 'Mera balance check karo please', it falls apart. Welcome to the code-mixing problem.
The Digital Personal Data Protection Act is now being enforced. Here's what your AI systems need to change - with specific technical requirements and implementati...
Everyone is building AI agents. Few are running them successfully in production. Here's what we've learned from deploying agentic systems across Indian enterprises.
Credit bureaus give you a number. Agentic AI gives you understanding. Here's how to build underwriting systems that see borrowers as economic actors in a connecte...
RBI has embedded AI governance expectations across multiple circulars covering IT governance, digital lending, and more. Here's what banks need to build - with sp...
Government AI projects claim sovereignty through data localization. But true sovereignty requires control over models, inference, and the entire AI supply chain. ...
