Tens of thousands of South Koreans had their financial data exposed online in recent weeks after at least nine major banks there suffered cyberattacks that suggested a dark future for online security.
One man, a 26-year-old living in southern China, was allegedly behind the attacks, according to an investigation by cybersecurity company CrowdStrike. Traces left online and failure to secure his own systems allowed investigators to piece together the man’s name, age, location and what he allegedly did.
The clues revealed that the devastating assault on some of Asia’s most sophisticated financial institutions was the work of an amateur. The attacks were launched with help from a few artificial intelligence chatbots, CrowdStrike found, before asking one to help sell the stolen data online and draft a fake résumé touting cybersecurity skills.
The incident was part of a wave of AI-enabled hacking that is advancing across the world, triggering alarm inside national governments as well as multinational companies. The technology can also help fix software bugs and enable defenders to move more quickly, but executives at AI and cybersecurity companies warn that they expect it to help malicious hackers more in the short term.
Last month, the U.S. National Security Agency advised websites to shut down if they could not immediately patch a major security hole in PHP, a programming language used by an estimated 70 percent of websites, discovered by a research firm using AI tools.
“Basically any pedestrian, non-hacker can take these frameworks and start leveraging them to go after different targets,” said Adam Meyers, senior vice president of counter adversary operations at CrowdStrike. The company has tracked an increase in the speed with which cyberattacks can move once inside a target’s networks, as seasoned hackers and rank amateurs alike get a boost from AI tools. “Anyone can be a super-hacker now,” Meyers said.
Concern began to spread earlier this year among cybersecurity experts and U.S. national security officials that the technology could trigger a chaotic era of widespread computer security breaches.
In April, after Claude chatbot maker Anthropic said its new Mythos AI model was too good at hacking to provide to anyone but trusted partners and the government, experts used it to find thousands of new computer vulnerabilities — some of which had lain dormant for decades.
The Trump administration, which had previously rebuffed calls to regulate AI, placed restrictions on how Anthropic and later OpenAI could release their newest models with the most potent cybersecurity skills. Only companies on a short list of government-approved firms can receive access to the top AI models without filters that limit them from helping with cybersecurity analysis, even for defensive purposes. (The Washington Post has a content partnership with OpenAI.)
The administration’s locked-down approach has drawn some criticism from business leaders and cyber experts, who say more people need access to top-tier models to protect themselves. When AI agents being trained inside OpenAI broke out onto the open internet and hacked into the AI company Hugging Face, the firm had to rely on free, or “open weights,” AI models to investigate the incident.
The inaccessibility of top models led some company leaders to think there was little they could do to respond to AI’s disruption of the security status quo, said Malte Ubl, chief technology officer of start-up Vercel, which builds AI tools for software developers. High-end AI models are now becoming easier to access, but the delay “cost the industry at least three, four, five months where things could have happened,” he said.
Many cyber experts still expect the security reckoning to intensify as AI tools that can find and exploit software bugs become more widely distributed and understood.
On Monday, The Post reported that cybersecurity researchers from the start-up Cathedral who set out to test Anthropic’s security found that they could access internal messages and meeting transcripts. Cathedral reported its findings to Anthropic, which said it resolved the newly discovered problems.
“I’m concerned this is what the future will look like,” Anthropic executive Logan Graham wrote in a post on X responding to The Post’s reporting. “Adversaries will use whatever models/tools they can. They’ll target everything with agents 24/7.”
Several major cybersecurity breaches and other incidents have been discovered in recent weeks, though not all have been confirmed to have depended on AI. Billions of dollars of digital assets were stolen from cryptocurrency exchange Binance last week. Security experts say it’s unlikely that any hacker is forgoing the use of AI tools right now.
Nick Carlsen, a senior investigator with TRM Labs, which tracks the movement of cryptocurrencies, said the company did not have specific evidence that the Binance hack involved AI. But the technology probably explains why the North Korean group behind the digital heist has recently appeared to be more powerful, he said.
“We have also seen a radical increase in sophistication. … AI is the logical explanation for these advancements,” he said.
Jason Clinton, Anthropic’s deputy chief information security officer warned in a post on X after the PHP vulnerability was discovered last month that more trouble lies ahead.
“As bad as things have gotten … they are going to get a lot worse before they get better,” Clinton wrote. Eventually, AI will help humans write code that has no security vulnerabilities, he said, but for now, the industry needs to “hold on.”
One reason to expect more challenges ahead is that free, open weight AI models, mostly made by Chinese companies, have rapidly improved. The best now perform at close to the same level as Anthropic’s Mythos when it was first released in April, Anthropic said in September.
“It was always super obvious that that capability was eventually going to be widely accessible,” said Will Wilson, chief executive of Antithesis, a start-up that uses AI to help companies make their code more secure.
“I think the idea that the frontier labs could keep this genie bottled up was always kind of laughable,” he said, referring to efforts by AI companies and the U.S. government to restrict access to AI models with top cybersecurity skills.
The hacker who broke into the South Korean banks had access to Anthropic’s Claude, but because the U.S. firm’s tools generally refuse to take illegal actions, free Chinese models were also used, including DeepSeek, CrowdStrike found.
“The models are interchangeable,” said Meyers, the CrowdStrike executive.
The U.S. government may need to step in to help certain sectors adapt to the AI onslaught, said Snehal Antani, chief executive of Horizon3, a cybersecurity company.
Critical infrastructure organizations like energy and water companies need to be “immediately and comprehensively” tested to figure out if there are bugs that could be exploited in their systems, he said. “We need resources to fund that remediation, and a public and private partnership to accelerate incident response.”
U.S. national security agencies have warned for years that companies that operate such infrastructure aren’t adequately prepared for the threat of cyberattacks that could disrupt supplies of water and energy. The FBI and Environmental Protection Agency said in August that “malicious cyber actors” have been remotely tampering with internet-connected equipment at water plants, causing loss of water pressure and flooding.
Veterans of the cybersecurity industry have warned for years that security holes are pervasive in all kinds of software. As society depends more on computer systems, the financial cost of frequent hacks and trying to defend against them has steadily increased. AI is now increasing the risks of weak security further, Wilson said.
“The industry as a whole has gotten by with this incredibly lackadaisical attitude toward security,” he said. “The bill is coming due in a really big way.”
The post ‘Anyone can be a super-hacker now’: The world faces a cybersecurity reckoning appeared first on Washington Post.
