From Copilot to Copirate: How data thieves could hijack Microsoft's chatbot

Microsoft patched a multi-stage Copilot exploit disclosed by researcher Johann Rehberger that combined prompt injection, automatic tool invocation, and ASCII smuggling to stealthily exfiltrate email contents and MFA codes; Rehberger published a proof-of-concept and Microsoft says mitigations were applied though details were not shared. The report warns enterprises building on Copilot/LLMs to address LLM-specific attack techniques and notes additional research and tooling demonstrating related risks.