About Lakera
Lakera sells AI Red Teaming, AI Agent Security and Workforce AI Security, which together help organizations secure employees, applications, and autonomous agents. Lakera states that its security prevents prompt injections, data leakage, and jailbreaks before they affect a business. Lakera directs its products at enterprise teams working with GenAI, agents, and MCPs. Lakera states that it is learning from 1M+ hackers around the world.[1][6][8]
AI Agent Security enforces security on every AI interaction an organization runs by inspecting what goes in, controlling what comes out, and governing what agents do. AI Agent Security provides visibility into agent usage and MCP-connected systems across an environment, including agents that teams did not explicitly build or register. AI Agent Security discovers AI agents and MCP-connected tools across an environment and identifies what data, APIs, and systems agents can access. Lakera says AI Agent Security protects agents across the full lifecycle, from discovery and risk assessment to runtime enforcement.[4][8]
AI Agent Security connects cloud, low-code, and custom environments. AI Agent Security assesses each agent based on its configuration, tools, data access, MCP servers, and autonomy. AI Agent Security enforces access controls and runtime guardrails to block threats, prevent data leakage, and govern live agent actions. AI Agent Security detects and blocks prompt attacks, data leakage, unsafe tool use, and unauthorized agent actions as agents operate. Lakera says AI Agent Security blocks indirect injection through connected tools before agents act on compromised instructions. AI Agent Security requires no changes to models or prompts. AI Agent Security extends the AI Defense Plane to the agents organizations build and deploy.[4][8]
AI Red Teaming gives teams a continuous workflow to evaluate, scan, and red team AI applications and agents. AI Red Teaming scoping maps AI architecture, workflows, tools, memory systems, and trust boundaries to build a targeted threat model and test plan. AI Red Teaming tests with prompt injection, jailbreaks, tool exploitation, and multi-turn attacks against real systems. AI Red Teaming provides exploit validation, severity scoring, and remediation guidance. AI Red Teaming is available for text, audio, image, and multi-modal systems. AI Red Teaming assessment covers goal hijacking, tool misuse, privilege escalation, indirect injection, and cascading failures. AI Red Teaming delivers validated vulnerabilities with reproduction steps, severity ratings, and affected workflows.[6]
Workforce AI Security discovers shadow AI, assesses risk, governs employee usage, and stops sensitive data exposure across AI apps, browser extensions, desktop agents, IDEs, and MCP-connected tools. Workforce AI Security covers the places employees prompt, paste, upload, and interact with AI day to day. Workforce AI Security extends visibility and policy enforcement from public chatbots to SaaS copilots to MCP-connected workflows. Workforce AI Security extends visibility into AI workflows where tools can access code, data, and enterprise systems.[1][7]
A Lakera report is based on a 30-day snapshot of production attack traffic observed by Lakera Guard. Lakera states that its security covers chatbots and audio bots across any model. Lakera states that policies can be customized to secure applications horizontally without changing code. Lakera describes Gandalf as an online game designed for learning about AI security.[1][10][11][12]
Dropbox uses AI Agent Security to help safeguard its LLM-powered applications. Lakera and Snyk have partnered on AI agent security research.[1][4][6][7][10]
Lakera on the record