Pangea

Pangea is an AI security platform, now presented alongside CrowdStrike, that helps security teams monitor and control AI activity across employee tool usage and in-house AI applications. It combines AI Detection and Response for workforce AI use with AI Application Guardrails for product teams.

What It Is

Pangea's AI Security Platform records AI usage with an audit trail and enforces runtime policy. Its collectors deploy via gateways, browser plugins, and a few lines of code in applications. Policies can allow or block GenAI use by application or model, as well as by prompt contents.

Threat Detection and Data Protection

The vendor states that it detects and blocks AI-specific attacks such as prompt injection, jailbreak attempts, and malicious content insertion in over 100 languages. It also claims 99% efficacy with roughly 5-40ms latency. For sensitive data, it describes runtime detection of 250+ types of confidential information and PII, with custom types supported. Rules can block, mask, hash, or apply format-preserving encryption.

Agents and Integrations

An MCP proxy lets teams wrap guardrails around agents and tools, with authentication and authorization for agent-to-tool and agent-to-agent communication. The platform also offers SIEM integration, and an AI gateway can apply consistent guardrails across models and applications.

Research and Services

Pangea Labs publishes AI security research, including the LegalPwn report, an interactive prompt injection taxonomy, and a prompt injection challenge. Pangea also offers AI red teaming and a product security workshop. The company states it holds SOC 2 Type 2, ISO 27001, and ISO 27701 compliance.