AUTOMATED VERIFICATION OF BRSR DISCLOSURES: A NEURO-SYMBOLIC NLP PIPELINE FOR OFFLINE GREENWASHING DETECTION, ETHICAL AUDITING, AND GREEN FINANCE INTEGRATION IN INDIA

The Securities and Exchange Board of India (SEBI) in 2021[9] rolled out the Business Responsibility and Sustainability Reporting (BRSR) mandate to adapt to inclusive growth and transition to a sustainable economy to mitigate the climate change impact. This caused Indian companies a massive paperwork headache, and everyone is racing towards the use of Large Language Models (LLMs). We have found three major drawbacks of this approach: AI tends to make up its own math (hallucinations) the energy needed to run these models is ironic given the green goal no Indian firm wants to leak sensitive data to a foreign cloud. This study breaks this cycle. We built a Neuro-Symbolic Pipeline designed specifically for environmental data in Principle 6[9]. By shrinking the model through 4-bit quantization, we made it light enough to run locally on a standard laptop, keeping the data private and the energy costs low. The real breakthrough is the deterministic regular expression (Regex) layer. Instead of just trusting the AIs feeling about a sentence, our code acts as a hard filter. It hunts for specific Indian metrics (Lakhs, Crores, Metric Tons) to determine whether a company is hitting its targets.