SenTeGuard provides real-time "idea security" for organizations deploying generative AI, intercepting and enforcing policy on LLM traffic before sensitive data leaks. The company also offers Moyo, a red-teaming service that maps how attackers could infer secrets by combining public sources, plus consulting for AI governance and generative AI optimization. Deployed as a proxy layer, SenTeGuard scans prompts and outputs in under 50ms with 150+ pre-built detection rules.
Funding
Funding not disclosed
Founders
Product
Problem
Traditional network security defends infrastructure but fails to protect the meaning flowing through AI interactions—classified data, trade secrets, and proprietary information can be leaked through LLM prompts and responses. Existing DLP tools only match exact patterns, missing semantic leaks like describing a trade secret in natural language or combining innocuous public data into sensitive conclusions.
Solution
SenTeGuard deploys as a real-time proxy between users and LLM endpoints, intercepting, analyzing, and enforcing policy on every AI interaction before damage occurs. The inspection engine scans inputs for PII, classified markers, prompt injections, and proprietary code patterns, then blocks, redacts, or flags content based on custom rulesets. Output validation checks responses for leaked system prompts or training data, while full audit logging captures every interaction for compliance reporting. Complementing the proxy, Moyo provides attack-path analysis for information reachability, mapping how public sources can be combined via LLMs to infer sensitive facts—treating inference itself as an exfiltration channel.
Target Audience
Primary customers are defense contractors handling CUI/classified data, financial services under SEC/FINRA obligations, healthcare organizations under HIPAA, and enterprises with 50+ employees using AI tools, as well as intelligence organizations and enterprises with sensitive IP needing inference exposure assessment.
Features
- Real-time proxy for all LLM API calls with under 50ms latency
- 150+ pre-built detection rules covering PII, secrets, code, and prompt injection patterns
- Custom policy builder with regex and semantic matching capabilities
- Output validation to prevent leakage of system prompts and training data
- 100% audit log retention with full context for compliance reporting (30 days to custom, per tier)
- Moyo reachability testing in black-box (external) and white-box (internal instrumentation) modes
- Pre-built policy templates for NIST AI RMF, EU AI Act, HIPAA, SOC 2, FedRAMP, CMMC, ITAR, and SEC rules
- Supports any HTTP-API model: OpenAI, Anthropic, Azure OpenAI, AWS Bedrock, Ollama, vLLM, and custom deployments