Lakera
4.3
Lakera is a security platform for generative AI with three products: workforce AI security governing how employees use AI including shadow AI discovery and prompt data protection, AI agent security providing runtime protection against prompt attacks and data leakage, and AI red teaming simulating direct and indirect attacks. It is API-first and cloud-native. Pricing is not published.
- Commercial use OK
- Enterprise contract
- Pricing
- Contact sales
- Free tier
- No
- Commercial use
- Permitted
- Japanese UI
- Not confirmed
- Enterprise contract
- Available
What it is good for
- Organisations with AI agents exposed to untrusted input
- Discovering which AI tools employees are already using
- Red teaming before an agent reaches production
- Security teams asked to sign off on an AI deployment
Strengths and trade-offs
Strengths
- Covers workforce usage and runtime protection, not just one
- Indirect prompt injection is explicitly in scope — many tools ignore it
- Red teaming as a product rather than a consulting engagement
Trade-offs
- No published pricing
- Adds latency to the request path
- Security tooling cannot compensate for a fundamentally unsafe design
Editors' review
Indirect injection is the one people miss
Direct prompt injection is well known. The attack that lands is indirect: a document in your RAG index containing instructions, retrieved and followed as though the user had typed them.
Covering that case is the right test when comparing products in this category.
Verdict
Worth evaluating once agents touch untrusted content. Ask specifically about indirect injection.
Compiled from official documentation and public sources, with our own editorial judgement. We do not publish figures we cannot source.