Skip to content
Submit incident
AII-1001-98597
ConfirmedA court, a regulator or the organization itself has stated the core facts, or the record cites a court or regulator source.Supporting sources are pending verification.

EchoLeak: Zero-Click Vulnerability in Microsoft 365 Copilot

Researchers discovered EchoLeak, a zero-click vulnerability in Microsoft 365 Copilot that could have allowed attackers to steal sensitive data via specially crafted emails without user interaction. Microsoft fixed the vulnerability server-side in May 2025 and confirmed no customers were impacted.

January 2025 Β· Microsoft Β· Security and cyberattacks
January 2025
AII-1001-98597
Quick facts
Category
Security and cyberattacks
Country
United States
Date
January 2025
Drafted by
Raidu Incident Detection

What happened

Researchers at Aim Labs discovered a zero-click vulnerability, dubbed EchoLeak, in Microsoft 365 Copilot in January 2025. The vulnerability allowed attackers to steal sensitive data from a user's environment by sending a specially crafted email containing hidden prompts in markdown syntax.

When users asked Copilot related questions, the system would retrieve the malicious email via its Retrieval-Augmented Generation engine. The hidden prompt would then instruct the AI to extract internal data and place it in links or images that automatically triggered outbound requests to attacker-controlled servers without user interaction.

Reported impact

The vulnerability could have exposed internal memos, strategic documents, and personal identifiers. Microsoft confirmed that no customers were impacted and there is no evidence the flaw was exploited in real-world attacks.

Index analysis

EchoLeak is described as the first-ever zero-click AI vulnerability affecting a large language model-based assistant. The attack exploited an LLM Scope Violation where the assistant could be manipulated into leaking information beyond its intended context, bypassing Copilot's Cross-Prompt Injection Attack defenses through reference-style markdown formats.

Governance controls

Microsoft fixed the vulnerability server-side in May 2025, requiring no action from users. The fix involved addressing how the assistant processes information from emails and documents when answering user questions.

Classification

Category
Security and cyberattacks
Industry
IT infrastructure and hosting
Country
United States
Harm status
Potential harm
Intent
Intentional
Timing
Post-deployment
Failure modes
Prompt injection
Consequences
None recorded
AI systems
Microsoft 365 Copilot

Sources and evidence

Sources not yet published

This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.

How sources are verifiedSuggest a source for this incident