Agents
Microsoft Entra Internet Access Prompt Injection Protection GA March 31: Network-Level AI Security Across 11 Model Providers
Microsoft's Entra Internet Access prompt injection protection goes GA on March 31, providing network-level blocking of adversarial prompts and jailbreak attempts before they reach AI models. Pre-configured extractors cover ChatGPT, Claude, Gemini, Grok, DeepSeek, Mistral, Perplexity, Cohere, Meta AI, Pi, and Qwen. This enforces consistent security across all generative AI applications without code changes. Notable architectural choice: security at the network layer rather than application layer means universal coverage regardless of which AI tools employees adopt.
↳ Follow the thread