# DeepKeep

> DeepKeep is an AI security platform that protects LLMs, AI agents, and vision models from adversarial attacks, prompt injection, jailbreaks, PII leakage, and hallucinations — across the full AI lifecycle from development through runtime.

## Products

- [DeepKeep for LLM](https://www.deepkeep.ai/llm): Security and trustworthiness controls for large language model deployments
- [DeepKeep for Vision](https://www.deepkeep.ai/computer-vision): Adversarial robustness and integrity protection for computer vision models

## Use Cases

- [AI Applications](https://www.deepkeep.ai/use-cases/deepkeep-for-ai-applications): Securing GenAI-powered customer-facing and internal applications
- [Employees](https://www.deepkeep.ai/use-cases/deepkeep-for-employees): Protecting organizations from risks introduced by employee use of AI tools
- [AI Agents](https://www.deepkeep.ai/use-cases/deepkeep-for-ai-agents): Security controls for autonomous and semi-autonomous AI agent systems

## Capabilities

- [AI Firewall](https://www.deepkeep.ai/capabilities/ai-firewall): Real-time runtime protection against prompt injection, jailbreaks, and PII/sensitive data leakage, with configurable block, redact, or alert enforcement
- [AI Red Teaming](https://www.deepkeep.ai/capabilities/ai-red-teaming): Automated and Vibe (human-steered) adversarial simulation across LLMs, computer vision, and multimodal systems to uncover exploitable vulnerabilities
- [AI Lens](https://www.deepkeep.ai/capabilities/ai-lens): Discovery of shadow AI usage across employees and developers, with role-based access control via IdP integration and built-in firewall protection, extending to developer tools like Cursor and Claude Code
- [AI Agent Scanner](https://www.deepkeep.ai/capabilities/ai-agent-scanner): End-to-end mapping of an AI agent's attack surface with prioritized, step-by-step remediation recommendations
- [Model Scanning](https://www.deepkeep.ai/capabilities/model-scanning): Static and dynamic AI supply-chain scanning for malware, CVEs, and tampering, with SBOM/MLBOM inventory before and during production

## Resources

- [Blog](https://www.deepkeep.ai/blog): AI security research, threat analysis, and practitioner guidance
- [Research](https://www.deepkeep.ai/resources/research): Published AI security research from DeepKeep
- [Glossary](https://www.deepkeep.ai/resources/ai-security-glossary): Definitions of AI security terms including prompt injection, jailbreaks, model poisoning, and adversarial attacks
- [Documentation](https://docs.deepkeep.ai/): Technical integration and platform documentation

## Company

- [About](https://www.deepkeep.ai/about)
- [Events](https://www.deepkeep.ai/events)
- [Partners](https://www.deepkeep.ai/partners)
- [Contact](https://www.deepkeep.ai/contact)
- [Book a Demo](https://www.deepkeep.ai/book-demo)