The Russian company Orion soft announced StarGuard AI — a platform for ensuring security when working with large language models.
StarGuard AI operates as a reverse proxy gateway, analyzing all traffic between users, corporate systems, and language models (LLMs). The developers emphasized that this allows companies not to block, but to securely integrate AI into workflows, controlling access, processed data, and received responses.
The platform's security system is multi-layered and uses its own ML models to detect confidential information, as well as detectors to prevent jailbreak and prompt injection attacks. This helps block toxic or prohibited content in AI chats, HR assistants, and support services, reducing reputational risks.
Another feature of StarGuard AI is centralized access management to various language models. Through a single gateway, both cloud services (ChatGPT, Claude, YandexGPT) and local LLMs can be connected.
The solution integrates with OpenWebUI, IDEs, and corporate platforms without the need to change client infrastructure.