Question 1

暗号文コンテキストインジェクションとは何ですか？

Accepted Answer

It is an attack, named by Adversa AI's Rony Utevsky, that hides malicious instructions as AES-256-GCM ciphertext so a model's static text filter cannot read them, then gets the model to decrypt and trust those instructions as its own code output.

Question 2

Grok攻撃はどのようなデータを窃取しますか？

Accepted Answer

The attack collects the victim's name, coarse location, subscription tier, and the full set of prompts in the conversation, then sends them to the attacker by appending them to a URL as parameters.

Question 3

xAIはこの脆弱性を修正しましたか？

Accepted Answer

No. Utevsky reported it to xAI on June 3, 2026 and followed up on August 4 and August 10, but as of August 19 the technique still worked on Grok.com and xAI had not given a mitigation timeline.

Adversa AI

Grokが暗号化されたウェブページのトリックでユーザーチャットを漏洩、xAIが警告してから11週間後

鋭い要約を。
毎日。

Adversa AI

Grokが暗号化されたウェブページのトリックでユーザーチャットを漏洩、xAIが警告してから11週間後

鋭い要約を。毎日。

鋭い要約を。
毎日。