Lakera Guard
used in
0
stacks
Commercial low-latency API for detecting prompt injection, data leakage, and unsafe content in LLM applications.
A hosted screening API that classifies prompts and model outputs for prompt attacks, PII, and content policy violations, configurable via policies. Detection models are continuously updated using threat data, including from Lakera's Gandalf prompt-injection game. Offers a free tier on its platform with paid and enterprise plans; Lakera was acquired by Check Point in 2025.
Strengths
- Purpose-built prompt injection detection served as a low-latency API
- Detectors continuously retrained on real-world attack data
- Policy-based configuration covering prompt attacks, PII, and content moderation
- SaaS and self-hosted deployment options for enterprise plans
Tradeoffs
- Closed source with unpublished pricing; quotes require contacting sales
- SaaS mode sends prompt and response data to a third-party API
- Post-acquisition integration into Check Point adds roadmap uncertainty