Case studies, deep dives and analyses on AI security, AI governance and the regulations reshaping both. Written by practitioners.
An AI sales intelligence platform with active GRC tooling and billion-dollar clients. We still exfiltrated 3.49 MB of cross-tenant enterprise intelligence in a single request.
An AI-native LMS lost a large enterprise deal over security. 28 vulnerabilities surfaced, 11 critical fixes shipped, and active deals progressed through review.
Traditional DLP, CASB, and UEBA tools were built for files and users - not AI runtime. Enforcing one simple policy requires four systems across three teams.
DPDPA enforcement and Shadow AI are converging into a high-probability, high-impact data risk surface for Indian organizations.
Enterprises are moving quickly with AI, but most lack visibility into where AI is used, what data it can access and what actions it can take at runtime.
Generative AI and agentic systems create whole new runtime surfaces. Enterprises are seeing attacks accelerate, while breaches involving shadow AI cost materially more.
Cybersecurity is shifting from reactive detection to predictive, AI-native, identity-centric and continuously governed systems. Static, scan-and-respond security will not scale into 2026.
Security teams are being asked to review AI projects before they go to production - and they-re being set up to fail.
Our BSides Bangalore 2026 paper: adversarial inputs aren't just misclassified - they're misclassified with high confidence. We hit 94% intent accuracy at 100ms P95, no inference-time LLM call.
A new paper from Google DeepMind proposes a paradigm shift: stop trying to make the model smarter and start making the architecture secure.
Treat MCP as a serious platform decision - schemas, runtime controls, ownership - or adopt it quickly, hit hidden costs and lose confidence in agentic systems altogether.