Bypassing Azure OpenAI’s Prompt Shield

Credit cards beware!

Pictured: Microsoft Copilot’s rendition of “a SQL query on a large monitor”…

This is a rework of a rework, but after playing around with some jailbreak prompts for another writeup, where AI wasn’t even the focus, it occurred to me that revisiting this might be worth it to really cement how easily malicious actors can abuse AI and LLMs.

Every AI tech company and their mom seems to be implementing AI within their tools, with Microsoft being …

azure azure openai bypassing copilot credit credit cards cybersecurity focus generative-ai-tools jailbreak large llm malicious malicious actors microsoft microsoft copilot monitor openai prompt prompt-engineering prompts query shield sql sql query writeup