OpenAI pitches Daybreak AI cyber defense to Korean CISOs after bank breaches

OpenAI pitched its Daybreak cyber defense package, which pairs its AI models with Codex and Codex Security, to about 60 Korean corporate security chiefs in Seoul on Oct 16, weeks after hackers breached four Korean lenders.

#What happened

OpenAI held a Cyber CXO Roundtable in Seoul's Gangnam district on October 16, 2026, briefing about 60 corporate executives, including chief information security officers, on countering AI-powered threats. Dominic Grillo, OpenAI's head of technical success, and security engineer Wana Tun told attendees that AI is rapidly honing its ability to unearth software flaws and map attack paths, leaving companies less time to respond, and advised defenders to give AI the code and operating context of their systems so it can plug holes before attackers strike. The briefing came weeks after hackers breached internet-facing support systems at four Korean lenders (Shinhan Bank, KB Kookmin Bank, Hana Bank, and BNK Busan Bank) in late September; investigators found traces of ARTEX, an open-source AI penetration-testing platform, on suspected attack servers, though how much AI was involved has not been confirmed. OpenAI showcased Daybreak, a defense package that links its AI models with Codex and Codex Security: the system finds flaws, tests in a controlled setting whether they can be exploited, and drafts fixes. OpenAI said many major Korean companies have adopted Daybreak or are considering it. "Companies must treat cybersecurity as a core management priority if they are to quickly identify and respond to major security risks in the AI era," said Kim Kyung-hoon, CEO of OpenAI Korea, adding that OpenAI will "help Korean companies bring AI into their existing security systems, see results on the ground and turn them into real defensive capability."

#Why it matters

The roundtable shows OpenAI turning the Codex Security line that debuted at DevDay 2026 into a managed defense package for enterprise CISOs, pitched where AI-augmented attacks are fresh news. It also marks the agent-security race going geographic: Korea's late-September bank breaches and Kakao VX's Oct 8 disclosure that data on about 34,000 Kakao Golf Booking customers had leaked gave OpenAI a live case study for Daybreak's pitch.