Lakera is a security platform for generative AI applications, agents and MCP integrations. It screens what goes into and comes out of developers' AI systems at runtime to block prompt injections, jailbreaks and data leakage, and also covers employee AI use and pre-release red teaming.

Key features:

Detect and block prompt attacks in real time with sub-50 ms latency.
Prevent sensitive data from leaking through prompts and responses.
Protect chatbots and audio bots across any model in more than 100 languages.
Set central security policies without changing application code.
Discover shadow AI use across apps and browsers and apply policies by user, app and action.
Run direct and indirect attack simulations with risk-based red teaming.

Lakera is an API-first, cloud-native service with enterprise integrations, and its threat data draws on more than a million hackers playing its security games. Developers can get started from the website, and enterprise plans go through sales; pricing is not published.