Indirect prompt injection against Microsoft 365 Copilot: a single email was enough
According to one report, an indirect prompt injection worked against Microsoft 365 Copilot: the attacker sent a single crafted email, and the user did not have to do anything. Another article reported that the input that allowed Copilot to be exploited had been made public.
Who is affected and how
- security leader
- Email and document inputs to AI assistants should be treated as an attack surface.
- developer
- Input handling in agents and assistants must not rely on untrusted content.
Sources
- Microsoft Copilot reveals secret input that allowed it to be hacked - Ars Technicaarstechnica.com · Aug 18, 2026
- Microsoft Copilot Hacked After Revealing Its Own Weaknesstheoutpost.ai · Aug 18, 2026
- Copilot reveals how it got hacked - Techzine Globaltechzine.eu · Aug 19, 2026
- Prompt Injection Attacks: How They Work & Defenses 2026deepstrike.io · Aug 20, 2026
- How prompt injection attacks work — and what to do about them | 1Password Community1password.community · Aug 20, 2026
- How do you prevent prompt injection attacks through MCP servers in 2026? | graftconcepts.comgraftconcepts.com · Aug 21, 2026
- Prompt Injection Attacks: How They Compromise AI Agents and MCP Tools in 2026 | MCP360mcp360.ai · Aug 22, 2026
- What is indirect prompt injection via MCP servers and how can founders protect their AI agents from this attack vector in 2026? | themercerclubnyc.comthemercerclubnyc.com · Aug 22, 2026
- How AI Guardrails Get Bypassed, Jailbreaks, Prompt Injection and 2026 Defensespasqualepillitteri.it · Aug 22, 2026
Machine-generated summary based on the titles and excerpts of the source articles. The facts come from the linked article. News monitoring and analysis, not security advice. Attributions belong to those who made them.
What do you make of this piece?
The split shows after 20 votes.
Why? Your answer is not shown publicly; it helps the editors.