Cybersecurity · 1 views
Researchers escape OpenAI Codex sandbox to run commands on host
Security researchers found two ways out of the OpenAI Codex sandbox, one of them capable of running commands on a developer's machine from Codex's most locked-down mode, with no approval prompt and nothing shown on screen.
AI Summary
Security researchers discovered two escape routes from OpenAI Codex’s sandbox. One flaw allowed the execution of commands on a developer’s machine while Codex was in its most restricted mode, bypassing any approval prompt and leaving no visible indication on the screen. The researchers reported both vulnerabilities to OpenAI on August 12. OpenAI addressed the issues and released fixes within eight days, according to Oren Yomtov of Accomplish AI.
AI summaries can be wrong sometimes—always verify important details using the source article.
How AI & Automation are usedMore from Cybersecurity
Continue reading recent Cybersecurity coverage
- Claude Opus 5 Helped Researchers Take Over OpenAI Staff Accounts via Chained FlawsContinue reading
- BragJack attacks hijack AI browser agents through malicious extensionsContinue reading
- TigerByte Cyber Emerges From Stealth With $3 Million in FundingContinue reading
- North Korean WaterPlum hackers infected 30,000 devices worldwideContinue reading
Support HappeningNow
Independent AI-powered news analysis is reader-supported. Your contribution helps cover infrastructure, summaries, and continued platform development.
Support HappeningNow