Secure workforce use of generative AI

Cloudflare’s SASE platform provides visibility and usage controls to safeguard how employees, contractors, and other users interact with GenAI tools like ChatGPT. For example, Cloudflare extends visibility and mitigates risks by classifying what AI tools are used, detecting misconfigurations, enforcing identity-based access policies, and restricting inputs and outputs of sensitive data.

Shadow AI refers to the unsanctioned use of AI tools without organizational approval, typically by the IT or security departments. Because Cloudflare’s SASE platform sees traffic between users and apps, organizations can identify and analyze use of these unsanctioned tools, evaluate the risks they pose, and enforce policies to allow, block, isolate, or enforce other protections.

AI security posture management (or AI-SPM) refers to a proactive approach to monitor and mitigate risks, including vulnerabilities, compliance issues, and data leaks. Cloudflare’s Cloud Access Security Broker (CASB) continuously monitors sanctioned GenAI tools via API to detect misconfigurations and exposed data at rest. Data Loss Prevention (DLP), combined with the Secure Web Gateway (SWG), provides real-time protection by discovering Shadow AI usage and actively inspecting user prompts and model responses. This enables organizations to prevent sensitive data leakage while interacting with AI models, enforce granular access control policies across all GenAI applications, and manage risks throughout the entire AI data lifecycle.

Prompt protection refers to how Cloudflare detects, blocks, and logs user inputs to GenAI tools to mitigate risks of data leaks and or malicious activity. For each user prompt, Cloudflare can analyze both the content and contexts of a prompt before enforcing a security policy. For example, if a prompt includes sensitive data like PII, source code, financials, or health information, administrators can set DLP policies to block and log that prompt. Cloudflare also detects user intent in AI prompts, preventing risky actions such as jailbreak attempts, requests for malicious code, or attempts to extract PII.