India is about to give itself the power to switch off your AI. That detail sat inside today's Economic Times article on India's proposed AI law. Most coverage skated past it. Here is what caught my eye: every mechanism India is floating already exists in the EU AI Act, live and enforced. Graded risk tiers. Light rules for chatbots and productivity tools, heavy rules for AI in banking, insurance, health, and critical infrastructure. Mandated safety testing. A national database for AI incidents. The power to disable dangerous systems in an emergency. Separate duties for the companies that build AI and the ones that deploy it. Europe wrote all of that down first. Which means you do not have to wait for India's draft bill to know what it will ask of you. The EU AI Act is your eighteen-month preview. The trap most Indian teams are about to hit: "we only call OpenAI's or Anthropic's API, so the model maker carries the risk." Both New Delhi and Brussels disagree. The rules split obligations between the provider who builds the model and the deployer who puts it to work. Run a system that screens loan applicants, triages patients, prices insurance, or ranks job candidates, and you are the deployer. Human oversight, telling people they are dealing with AI, keeping logs: that sits with you, whatever model runs underneath. The reassuring headline says chatbots count as low risk. True on their own. Drop the same chatbot into a lending or diagnostic workflow and the context reclassifies it. The comfort disappears. So start now, against a live regime instead of a rumoured one. Inventory every AI system you run. Classify each by how it gets used, not what it is. Work out whether you are the provider or the deployer. Map your gaps before a regulator or a customer's procurement team asks you to prove it. This is the pattern worth naming. A dozen national AI regimes are arriving, and most of them rhyme. Audit for one and you have most of your answer for the rest. India just proved the rhyming has started. Full breakdown here 👇 linkedin.com/pulse/indias-ai…