How did the Heapjack and Overpatch vulnerabilities discovered by Accomplish AI researchers let a malicious repository escape OpenAI Codex’sIllustration of the security boundary examined in the Heapjack and Overpatch research.
AI 提示詞
Create a landscape editorial hero image for this Studio Global article: How did the Heapjack and Overpatch vulnerabilities discovered by Accomplish AI researchers let a malicious repository escape OpenAI Codex’s. Article summary: Accomplish AI found two ways for content in an untrusted repository to cross Codex’s sandbox boundary. Heapjack could lead to unsandboxed command execution even in read-only mode; Overpatch could make Codex write outside. Topic tags: general, general web. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fake numbers, click
openai.com
檢查別人的程式碼儲存庫,未必只是「讀取檔案」。Accomplish AI 發現,攻擊者可利用儲存庫中的內容,讓 OpenAI Codex 的高權限輔助元件越過原本的沙箱限制。兩項漏洞的結果並不相同:Heapjack 可從只讀模式觸及開發者的主機並執行命令;Overpatch 則可突破工作目錄的寫入範圍,而且不會跳出核准提示。26