Impart Security, Lakera, Lasso Security and Prompt Security

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]}

About Lasso Security

Lasso Security sells an AI Security Platform that covers discovery, AI risk assessment and runtime protection. Lasso Security offers protection for every AI agent an organization builds and deploys, wherever it runs. Lasso Security secures autonomous AI agents, LLM applications and agentic workflows from build-time through runtime. Lasso Security states that decentralized ownership has led to Shadow AI across the enterprise. Lasso Security states that security teams have zero visibility into the AI tools and agents that employees are using and creating. Lasso Security states that everyone from HR to Marketing is now using or building agents.^[1][4][5][6]

Lasso discovers every agent and AI application across an environment, consolidates them into a single inventory and assesses risk across the agentic ecosystem. Lasso inventories AI components including models, MCP servers, delegated agents, databases, third-party tool connectors, identity boundaries and authorization policies. Lasso connects through native integrations with CI platforms and low code no code agent builder platforms such as Amazon Bedrock, Azure AI Foundry, Microsoft Copilot Studio, Google Vertex AI and Salesforce Agentforce. Lasso uses an agentless, read-only integration into CI pipelines and native integrations with cloud providers and third-party agent builder tools.^[4][5]

For posture management, Lasso visualizes how every agent is built, what is exploitable and how to fix it before the agent reaches production. Lasso posture coverage spans OWASP, MITRE, NIST, the EU AI Act and internal policies. Lasso identifies misconfigurations, risky tool access, weak guardrails, exposed dependencies and potential attack paths. Lasso assesses agentic architecture for compliance and governance requirements, including NIST, the EU AI Act and company-specific AI policies.^[7]

Lasso identifies exploitable vulnerabilities through high-agency adversarial testing and auto-updates policies with purple teaming. Lasso integrates with a CI/CD pipeline so red teaming simulations can run automatically on every application update. Before an attack runs, Lasso performs optional reconnaissance to map the profile of the application. The Lasso agentic attack engine runs multi-turn adversarial simulations that evolve across conversation turns. Lasso red teaming runs from the Lasso tenant, with no agents, no code changes and no source code access.^[1][4][8]

Lasso enforces policies inline at the proxy, API or AI Gateway layer. Lasso detects and terminates AI threats across an environment, with coverage including MITRE and OWASP. Lasso monitors every interaction across the AI stack, from tool calls and RAG retrievals to sub-agent communications. Lasso enforces predefined and custom security policies inline before agent actions execute. Lasso builds a behavioral baseline for every AI user and agent in an environment. Lasso flags deviations from the behavioral baseline in real time, including novel attack techniques with no existing signature. Lasso runs as a lightweight layer alongside an existing stack with no architectural changes required. Lasso shows alerts and audit logs directly in Lasso or streams findings into a SIEM alongside other security telemetry.^[1][4][9]

Lasso's Intent Deputy secures AI agents with real-time intent analysis. Lasso takes action according to a security threshold, either blocking the request or sending an alert. Lasso decodes the semantic meaning behind every request to detect hidden malicious commands behind techniques like obfuscation, adversarial embeddings and payload splitting. Lasso uses an 8-dimensional coordinate system to provide human-readable evidence for legal and risk teams.^[10]

Lasso enables enterprises to use AI across every department, from SaaS and web chatbots to local desktop agents and MCPs. Lasso discovers, inventories and assesses every AI asset across an enterprise, from models to SaaS and web chatbots to local desktop agents or homegrown agents and their connected tools. Lasso detects in real time whether PII, PCI or other sensitive data is shared in prompts or file uploads, and masks that content before the prompt is sent to third-party models. Lasso limits which AI agents are available to certain roles and restricts high-risk actions based on clearance levels. Lasso maps policies to NIST, OWASP and MITRE using plug-and-play templates, or enforces internal rules with custom policies.^[6]

Lasso Security governs the lifecycle of AI agents from build-time to runtime for agents built on Vertex AI, Microsoft Copilot, AWS Bedrock, Salesforce Agentforce or any other cloud or third-party service. Lasso gives risk scores that highlight which agents require immediate attention. Lasso maps agents and their connections to third-party tools, databases and internal services to identify potential attack paths. Lasso applies real-time guardrails at the moment of interaction that block malicious inputs, sanitize outputs and enforce content policies.^{[8][11][12][13]}

About Prompt Security

Prompt Security offers protection for both employee use of AI tools and the apps an organization builds. Prompt Security continuously tracks prompts per day in organizations of 500 to 1,000 employees.^[4][5]

Prompt Security holds that securing agentic AI is fundamentally different from securing traditional software. Prompt Security states that agents act autonomously, call tools and reason over untrusted input, so prompt injection, identity and runtime behavior have to be governed in real time and not only at the perimeter.

Prompt Security provides a security skill suite for OpenClaw, NanoClaw and Hermes agents.^[6]

Prompt Security provides a tool that tracks and analyzes OpenClaw deployments across an organization. Prompt Security provides a tool that tests and hardens a system prompt against dynamic LLM-based attacks.^[4][9]

Prompt Security publishes news, research, tools and analysis for security teams responding to AI threats. Prompt Security describes an AI gateway as a centralized layer between an organization's users or applications and the AI models or services they call. An AI gateway, as Prompt Security describes it, lets security teams monitor, control and secure that traffic from one place instead of instrumenting every application separately.^[4]

Prompt Security curates an interactive, continuously updated directory of 471 companies building security for agentic AI, organized across 14 functional categories. Both pure-play startups and dedicated AI-security lines from larger vendors are in scope for the Prompt Security directory. Entries in the Prompt Security directory are compiled from vendor websites, product documentation, funding announcements, public company registries and analyst coverage. Placement in the Prompt Security directory reflects publicly stated capabilities, not vendor marketing tiers. Category definitions and threat coverage in the Prompt Security directory are mapped against established, independent frameworks for AI and LLM security. The full dataset behind the Prompt Security directory is open and machine-readable.^{[2][4][5][6][9]}