Researchers escaped OpenAI's Codex sandbox two ways, one running commands on a developer's machine from its most locked-down mode. OpenAI has patched both.
OpenAI fixed two Codex sandbox escape vulnerabilities after researchers showed how malicious code could bypass key security restrictions.
"BragJack" has been released, a method for hijacking AI assistants running within browsers—such as Chrome, Edge, Perplexity ...
Researchers escaped OpenAI's Codex sandbox two ways, one running commands on a developer's machine from its most locked-down mode. OpenAI has patched both. OpenAI admits it did not disclose an ...