News, research & announcements

Product launches, funding, analyst recognition and threat research from the Kavach-AI team. For press enquiries, reach press@kavach-ai.com.

Featured

Kavach-AI raises $38M to secure the agentic era

Led by Northgate Ventures, the seed and Series A will fund the Inside-Out platform, double the detection-research team and expand coverage for MCP servers and multi-cloud AI workloads.

kavach.press --latest
FUNDING$38M seed + Series A
TEAMheadcount 34 → 60 in 2026
COVERAGEMCP · AWS · Azure · GCP
RESEARCHdeeper prompt & tool-abuse detection
Latest

From the newsroom

Product

MCP Security guardrails reach general availability

Call-time blocking, wildcard tool scoping and compliance-ready audit logging are now available to every Kavach-AI customer.

October 2026
Research

Prompt-injection campaign targets exposed MCP servers

Our threat-research team documents a wave of attempts to hijack agent tool calls and walk off with cloud credentials.

September 2026
Recognition

Kavach-AI named a Leader in the 2026 AI Security Analyst Radar

Recognized for the strength of its live inventory, attack-path scoring and real-time D&R for agentic workloads.

August 2026
Company

Kavach-AI opens a London office for EMEA

Expanding support and solutions engineering for customers across the UK and Europe.

June 2026
Open source

kavach-mcp crosses 10k monthly downloads

The open-source detection middleware that powers the platform reaches a new milestone, with community-contributed rules.

April 2026
Product

AI Data Protection adds format-preserving tokenization

Protect regulated fields before inference while analytics and development keep using the protected values.

February 2026
Press kit

Media resources & contact

Logos, product imagery and boilerplate for coverage. Interviews on request.

Brand assets Product screenshots Company boilerplate press@kavach-ai.com
Stay in the loop

Threat research in your inbox.

See how Kavach detects prompt injection, tool abuse and data exfiltration — on your own stack.