Solar Group has released version 8.4 of its Solar Dozor DLP system. The update introduces tools for monitoring and restricting data transmission via public AI services, including ChatGPT, Gemini, Claude, DeepSeek, and Perplexity, as well as Russian corporate messengers.
According to Solar Group's research from the first half of 2026, 40% of queries to public AI services contain sensitive information. Specifically, 41% of transmitted data consists of source code and configurations, 30% includes personal and financial records, 18% is intellectual property, and 11% comprises passwords and API keys.
Solar Dozor 8.4 analyzes prompts and files on workstations before they are sent to AI services. Depending on security policies, the system operates in monitoring or blocking modes. It also saves dialogue history, allowing security teams to review the context of potential incidents and refine organizational policies.
The update enhances document recognition capabilities by adding a template for identifying military ID copies. This feature helps prevent unauthorized sharing of sensitive HR documents. Additionally, the system supports monitoring on Windows, Linux, and macOS, with new self-protection mechanisms for macOS agents requiring a PIN code for modifications.
To improve performance, version 8.4 introduces load balancing between servers and asynchronous processing for long-running operations. The web interface has been updated with new filtering options for events and UBA modules, while the email approval process now allows for batch processing of outgoing messages.