North Korea's Kimsuky hacking group has built its own artificial intelligence tooling to support cyberattacks, according to a report summarized by Yahoo News Canada under the headline "North Korean hacking group builds AI tools for cyberattacks, report says."
A separate account from Chosunbiz goes further on the specifics, reporting that Kimsuky has constructed a local large language model — an AI system run on its own infrastructure rather than accessed through a commercial provider — in order to automate AI-driven cyberattacks aimed at South Korea.
The detail that matters most in those two short reports is the word "local." Commercial AI services from major vendors sit behind usage policies, account checks, and abuse monitoring, and state-linked groups that lean on them leave a trail. A model an attacker hosts itself has no such gatekeeper. Neither source item available here describes how capable Kimsuky's model is, what data trained it, or which specific attacks it has been used in, so those questions remain open.
Kimsuky is identified in both reports as a North Korean group, and Chosunbiz frames South Korea as the target of the automated campaigns.
Why it matters: if a state-backed group can run its own AI in-house, the industry's main lever for restraining misuse — policing access to the big commercial models — stops working, and the cost of mounting large volumes of tailored attacks falls sharply.