AI Is Cutting the Time to Find a Hack From Weeks to Hours. Here Is What Security Teams Should Do About It.
South Korean hacker-turned-CEO Park Chan-am told a government security forum that artificial intelligence has shattered old assumptions about how fast criminals break into corporate systems, and that most organisations aren't ready.
Key points
- Park Chan-am, CEO of security firm Steelion and a former adviser to South Korea's National Police Agency, told a government audience in 2026 that AI has cut the time needed to find exploitable software flaws from four weeks to under one day.
- Over 300,000 servers running local AI testing software were confirmed exposed to the open internet in 2026, up from roughly 1,000 when the same vulnerability was first flagged in 2024.
- Park argued that traditional vulnerability scoring systems are no longer enough, and that security teams should instead prioritise flaws already being used in real criminal attacks.
- Park warned that AI agents, software programs that carry out tasks automatically, can leak sensitive data if given too many system permissions.
- He concluded that no technology replaces human judgement in confirming a real threat.
Park Chan-am was breaking into computer systems competitively at age eleven. He went on to advise South Korean government agencies, helped defend the country against cyberattacks linked to North Korea, and now runs Steelion. This month, as CSO Online reported, he addressed a forum organised around South Korea's fifteenth national Information Security Day. His message was blunt: AI has changed the rules, and most organisations haven't caught up.
How much faster can criminals actually work now?
Park put a number on it. Finding a usable weakness in a company's software used to take a skilled attacker at least four weeks. With AI tools, the same job now takes less than a day. A company that once had weeks to notice a problem and fix it may now have hours.
This changes what counts as a target. Every piece of software a business runs is now a more attractive entry point than it was before. Park's framing on supply chain accountability is worth noting: once software is installed on a company's systems, he argued, it stops being a vendor problem and becomes the company's own.
Park also flagged a rising blind spot in local AI testing. When staff experiment with open-source AI models on office servers, they often leave those servers exposed to the internet by accident. When the remote code execution flaw in Ollama, a popular tool for running AI models locally, was discovered in 2024, roughly 1,000 servers were exposed. By 2026, that figure had grown to over 300,000. As we reported on 17 July, the NadMesh botnet is already scanning for exactly these exposed servers, harvesting cloud credentials from machines left open this way.
Should you worry about AI agents leaking your data?
For companies using AI agents, programs that read emails, analyse documents and connect internal systems automatically, Park raised a specific concern about permissions. If an agent has access to more systems than it needs, a criminal who tricks it with a poisoned document can use that access to pull out confidential data. Stale permissions left behind by former employees or contractors make this worse, because AI links those overlooked access rights together automatically. The fix, Park said: design what the AI is allowed to touch before deciding what it will do.
How should teams decide which flaws to fix first?
Park pushed back against relying solely on the Common Vulnerability Scoring System (CVSS), the standard scale security teams use to rank how dangerous a flaw is. A CVSS score tells you how bad a flaw looks on paper; it doesn't tell you whether criminals are actively exploiting it right now. He suggested pairing it with the Exploit Prediction Scoring System (EPSS), which estimates the real-world probability of exploitation, and the US government's Known Exploited Vulnerabilities (KEV) list. Our coverage of two recent CISA alerts shows what that looks like in practice: flaws already weaponised in the wild get a hard deadline regardless of their CVSS rating.
A flaw with a CVSS score of 7.5 looks unurgent. The same flaw, if it's on the KEV list and its EPSS probability has jumped from one percent to ninety percent in two months, is on fire. Patch it first.
The real test for any security team isn't whether their tools can scan broadly. It's whether the people running them can cut through the noise and decide what matters today.
