PromptArmor
Tests generative AI vendors for security risks like indirect prompt injection across 26 risk vectors.
About
PromptArmor is a San Francisco-based AI security startup co-founded by Shankar Krishnan, built on the premise that all net-new AI risk is third-party risk. Its platform scans vendors for newly added AI features and policy changes that are hard to track, and tests generative AI vendors for risks such as indirect prompt injection. It checks 26 risk vectors mapped to frameworks including the OWASP LLM Top 10, MITRE ATLAS, and the NIST AI Risk Management Framework.
The company's researchers demonstrated real-world prompt injection issues, including a 2024 disclosure showing Slack's AI summarization could be manipulated to leak private channel content. Note: several unrelated open-source projects on GitHub also use the name PromptArmor; this entry covers the commercial security startup, not those repos.
What it can do
- Third-party AI vendor risk assessment and monitoring
- Indirect prompt injection testing across 26 risk vectors
- Detection and removal of injected prompts from data samples
- Risk reports mapped to OWASP LLM Top 10, MITRE ATLAS, and NIST AI RMF
- Continuous monitoring of vendor AI features and policy changes
Setup
Sign up on the PromptArmor website to submit a vendor for assessment or request API access for continuous monitoring.
Prerequisites
A list of AI vendors or third-party AI features to assess.
Use cases
Works with
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 94; Trending score 70/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.