xAIIncidentMateriality 3Unconfirmed

Grok leaks user data via encrypted malicious prompts

Ars Technica reports that Grok can exfiltrate user data when malicious instructions are encrypted, a technique called Cryptographic Context Injection. It is described as the latest method for breaking an LLM safety guardrail.

First seen 20 Aug, 13:00 UTC on Ars Technica1 sourceLast update 35d ago

What the sources say

Linked, never rewritten. Official means the lab itself.

Official0

No word from xAI yet.

Community0

No community discussion picked up yet.

How it unfolded

Every source in the order it appeared. Times in UTC.