AI tools lower the barrier for hackers, forcing banks and agencies to adapt

Cybersecurity firm CrowdStrike identified a 26-year-old man in southern China as the likely operator behind a series of cyberattacks on at least nine major South Korean banks. The breaches, which exposed financial data for tens of thousands of customers, demonstrated how artificial intelligence tools are transforming the capabilities of individual attackers.

The investigation revealed that the suspect used several AI chatbots to launch the assault, sell the stolen data online, and draft a fake resume listing cybersecurity skills. Traces left in the digital environment allowed investigators to determine his name, age, and location. The finding underscores the shift toward automated and AI-assisted hacking, a trend that has triggered alarm among national governments and multinational corporations.

Adam Meyers, senior vice president of counter adversary operations at CrowdStrike, stated that the speed of modern cyberattacks has increased as both seasoned professionals and amateurs utilize these tools. He noted that basic frameworks allow non-experts to target sophisticated systems. "Anyone can be a super-hacker now," Meyers said. The company has tracked a rapid increase in the pace at which attackers move once inside target networks.

The U.S. National Security Agency recently advised websites to shut down if they could not immediately patch a major security hole in PHP. A research firm used AI tools to discover the vulnerability in the programming language, which is estimated to be used by 70 percent of websites. This incident highlighted the dual nature of AI, which can help fix software bugs but also empowers malicious actors to find and exploit weaknesses more quickly.

In April, experts used the Mythos AI model from Anthropic to identify thousands of new computer vulnerabilities, some of which had remained dormant for decades. Following this, the Trump administration placed restrictions on how Anthropic and OpenAI could release their newest models with potent cybersecurity skills. Only government-approved firms can currently access the top AI models without filters that limit their cybersecurity analysis capabilities.

Malte Ubl, chief technology officer of start-up Vercel, argued that the restriction on access to top-tier models cost the industry at least three to five months. He noted that when AI agents from OpenAI hacked into the AI company Hugging Face, the firm had to rely on free, open-weight models to investigate. "I think the idea that the frontier labs could keep this genie bottled up was always kind of laughable," said Will Wilson, chief executive of Antithesis.

Last week, billions of dollars in digital assets were stolen from cryptocurrency exchange Binance. Nick Carlsen, a senior investigator with TRM Labs, said while there is no specific evidence AI was used in the heist, the technology likely explains the recent increase in sophistication by the North Korean group behind the attack. "We have also seen a radical increase in sophistication.. AI is the logical explanation for these advancements," Carlsen said.

Jason Clinton, deputy chief information security officer at Anthropic, warned that security challenges will likely worsen before they improve. He stated that while AI will eventually help write code with no security vulnerabilities, the industry must hold on for now. The rapid improvement of free, open-weight AI models, many from Chinese companies, is expected to make advanced hacking capabilities widely accessible, increasing the pressure on critical infrastructure to adapt.