Generative AI Security: Are Your Developers Pasting Secrets Into LLMs?

Описание и применение

Внедрение GitGuardian в AI‑шлюз позволяет обнаруживать и блокировать утечки секретов в запросах к LLM, уменьшая риск раскрытия данных.

Задача

Developers inadvertently paste secrets into AI prompts or tool calls, leaking them to third‑party model providers without any repository involvement.

Результат

Deploying a GitGuardian‑backed AI gateway surfaced hundreds of secret incidents within weeks, with many valid findings that never touched a git repository.

Рабочий процесс

• 1. Configure GitGuardian with Custom Sources and a service account. 2. Deploy an internal AI gateway that forwards each request to GitGuardian’s /v1/scan/create-incidents API with the Custom Source UUID. 3. GitGuardian scans the payload in memory using 600+ detectors and creates incidents in the dashboard. 4. The gateway either forwards the request (non‑blocking) or rejects it (blocking) based on the chosen mode.

Доказательства

GitGuardian detected over 1.27 million leaked secrets tied to AI services, an 81% jump from the previous year.

Источник

One customer deployed GitGuardian behind their internal AI gateway and surfaced hundreds of secret incidents within weeks.

Источник

The AI gateway forwards the payload to GitGuardian's scan API, tagged with your Custom Source UUID. We scan in memory and return findings.

Источник