DeepKeep is a model-agnostic AI security platform for organisations that want to protect AI applications, AI agents and models from development through to production use. It includes an AI firewall for runtime detection and guardrails, AI red teaming to test robustness, AI Lens for visibility and access control, a scanner that maps what an agentic workflow can reach, and model scanning for supply-chain checks. Separate offerings cover language models and computer vision. According to the vendor, the platform can run as SaaS, on-premises, in the customer's own VPC or in an air-gapped environment.
DeepKeep is an Israeli company, which places it outside the EU and the EEA. A European organisation that chooses the SaaS option processes data with a provider under Israeli jurisdiction. The platform itself is not open source. The on-premises, own-VPC and air-gapped options the vendor describes give more control over where data and models stay, but the software and its support still come from a non-European supplier.