Lakera
The AI-native security platform to accelerate GenAI, now part of Check Point
Overview
Lakera built genuine credibility in the LLM security space through Gandalf, its viral prompt-injection game that crowdsourced over a million real adversarial attempts, and turned that data into one of the more mature prompt-injection detection engines on the market. Its sub-50ms latency claim matters in practice: a lot of runtime security tools add noticeable lag to production LLM apps, and Lakera's engineering has clearly been optimized to avoid that. It's a solid choice for engineering teams that want lightweight, fast guardrails rather than a full compliance suite.
The big change to watch is the Check Point acquisition (~$300M, closed September 2025). The product is still live and actively positioned as the runtime prompt-injection and PII layer within Check Point's broader security portfolio, but buyers should expect pricing, support, and roadmap decisions to increasingly reflect Check Point's enterprise sales motion rather than an independent startup's. Pricing remains largely custom and undisclosed beyond a rough community/paid split, so get a real quote before assuming cost.
Key Features
Lakera Guard: Real-time prompt injection and jailbreak detection with sub-50ms runtime latency for production AI applications.
AI Agent Security: Runtime monitoring and protection specifically for autonomous AI agents, not just single-turn chatbots.
PII & Data Leakage Protection: Detects and blocks sensitive data exposure in both prompts and model outputs.
AI Red Teaming: Risk-based adversarial testing to uncover vulnerabilities before an AI application goes live.
Multilingual, Multimodal Coverage: Threat detection across 100+ languages and multiple input modalities.
Workforce AI Monitoring: Visibility into and control over employee use of third-party AI tools.
Pricing
Starting price
Free community tier; paid plans reportedly start around $99/mo, enterprise pricing custom
Community: Free tier to get started with core prompt injection detection at limited volume.
Paid Plans: Reportedly start around $99/month for higher volumes and additional features.
Enterprise: Custom, per-request pricing negotiated through Check Point's sales channel.
Disclaimer: pricing may change, confirm on Lakera's own pricing page before buying.
Pros
Low runtime latency: Sub-50ms detection makes it practical to run in production without degrading user experience.
Battle-tested detection data: Built on real-world adversarial data from Gandalf's 1M+ crowdsourced attack attempts.
Broad language and modality coverage: Detects threats across 100+ languages and multiple input types, not just English text.
Now backed by Check Point: Acquisition brings enterprise distribution and resources beyond a standalone startup.
Cons
Post-acquisition transition: Sales, support, and roadmap now route through Check Point, which may change the product's independent identity over time.
Largely undisclosed pricing: Beyond a rough free/paid split, enterprise pricing requires a custom quote.
Point solution for runtime/red-teaming: Doesn't cover model supply chain scanning the way Protect AI or HiddenLayer do; may need to pair with another tool for full lifecycle coverage.
What Makes It Unique
Crowdsourced adversarial data via Gandalf: Lakera's detection models are trained on real-world prompt injection attempts from a viral public game with over a million participants, giving it a data advantage few competitors can replicate.
Kay Score
7.4
/ 10
Tool Information
Pricing
Free community tier; paid plans reportedly start around $99/mo, enterprise pricing custom
Category
Automation & AI Agents
Platform
Web / iOS / Android
Last Updated
