India proposes AI audit rules that could increase scrutiny of OpenAI and Anthropic
India’s cybersecurity agency CERT‑In has drafted Artificial Intelligence Security Auditing Guidelines that would subject AI systems to a risk‑based review. The draft classifies AI tools as low, moderate, high or critical risk and places special focus on autonomous agents, requiring them to run under the principle of least privilege and, for high‑risk agents, to include a “kill switch.”
Key points
- CERT‑In proposes risk‑based AI audit guidelines covering low to critical risk categories
- High‑risk autonomous agents must use a “kill switch” and operate under least‑privilege principles
- AI incidents would need to be reported to CERT‑In within six hours
The proposal also calls for continuous monitoring of critical‑risk systems, independent security audits by CERT‑In‑empanelled firms, and a six‑hour reporting window for AI‑related incidents. Companies would need to document data provenance, bias mitigation, and safeguards throughout model development, and label or watermark synthetic content where appropriate. While the guidelines are not yet law, they signal growing regulatory attention on AI safety and could affect how OpenAI, Anthropic and other providers design and deploy their models in India.
If adopted, the rules would add compliance overhead for AI developers and may shape global best practices for AI security audits, especially for agentic systems that can act autonomously across tools and data sources.
India’s new AI security audit rules could put OpenAI, Anthropic under greater scrutiny
cnbctv18.com · 4 November 2026
Loading the full article…
This text was published by cnbctv18.com and written by Ashmit Kumar. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Policy & Regulation
All →- UK found to be dangerously dependent on China for critical minerals · 1 src
- arXiv study finds end-users rate AI transparency as less useful than builders and policymakers · 1 src
- Anthropic warns IPO investors it may imminently hit AI self-improvement threshold · 9 src
- Google says schools must accept AI features to use Search starting next year · 35 src
- Anthropic’s IPO prospectus outlines unusual governance and succession rules for its founders · 2 src
Comments
via GitHub Discussions