SteadyDetected Oct 9

CrowdStrike finds possible bank hacker's CV among exposed AI logs

Steady
1.0 momentum
PressHacker NewsReddit
3 stories across sources

What's happening

CrowdStrike says investigators found an exposed set of AI interaction logs that include what looks like a hacker's CV and show use of Claude Code and the agentic penetration tool ARTEX in attacks on South Korean banks. Reports note the actor used a combined stack that included ARTEX, DeepSeek v4.1-Flash, GLM-5.3, Grok 4.6, and Claude Code. South Korean officials say AI agents appear to have been used to hack some of the country's biggest banks last week, and the logs suggest the suspected actor may be a Chinese speaker.

Why it's trending

Because investigators recovered AI conversation logs and tooling details that link agentic open-source models to a recent, high-profile bank intrusion in South Korea.

SignalHolding at its usual pace, confirmed across 3 independent source types.

Story volume

Stories per day
10-0610-08

Angles you could write

contrarian take

If finding a hacker's CV in AI logs doesn't make you rethink how you store agent conversations, nothing will, this is on us, not the models.

+2 more angles for this topic with an account β€” all it takes is your email.

More rising in AI & Tech

All rising AI & Tech trends β†’