https://blog.calif.io/p/codex-hacked-a-samsung-tv
Article
-
OpenAI’s Codex agent was used to find and exploit a vulnerability in a Samsung TV
-
Firmware source code was provided, letting the AI identify attack surfaces directly
-
Exploit chain traced through Novatek chipset drivers Samsung had shipped in the TV
-
Demonstrates agentic AI doing end-to-end security research with minimal human steering
Discussion
-
Commenters debate credit: did Codex hack it, or did the human wielding Codex?
-
Samsung TVs have been highly hackable for a decade — bar may not be high
-
Providing firmware source is seen as the key enabler, not raw AI capability
-
Others note Codex is inconsistent — impressive in one session, useless in another
Discuss on HN