Blog
Chinese AI Tool Artex Linked to Hacks in Japan and South Korea
22 October 2026
Artex is a Chinese open-source AI penetration-testing tool. South Korea’s government says it was highly likely used in data breaches at more than seven financial institutions. Japan is dealing with a similar wave of attacks, but its police say it isn’t yet clear whether the cases are linked. The tool’s developer has since shut down public access.
AI-powered cyberattacks are no longer a theoretical risk. A tool built to find security flaws now sits at the centre of breaches across East Asia, and the story raises hard questions about open-source software, AI misuse and who is responsible when a defensive tool turns offensive.
What Is Artex?
Artex is a Chinese-developed, open-source penetration testing tool. According to CrowdStrike’s analysis, it connects to large language models to carry out autonomous hacking. Penetration testing is normally a defensive practice: security teams simulate attacks to find weak points before criminals do. Artex automated much of that work using AI.
The same automation that makes it useful to defenders makes it attractive to attackers. A tool that scans for weaknesses and acts on them with little human input lowers the skill needed to run a serious attack.
The South Korea Breaches
South Korea’s government said it was “highly likely” the tool had been used in data breaches at more than seven financial institutions, including major banks. The Financial Services Commission said more than 68,000 people had been affected.
One example: Shinhan Bank said loan-application information for around 25,000 customers had leaked, including names, phone numbers and annual income. Details like these are exactly what fraudsters use for phishing, impersonation and identity theft.
The Japan Connection
In Japan, around 20 companies have said their data may have been compromised in a spate of similar cyberattacks potentially affecting millions of customers.
The picture here is less settled than in South Korea. Japan’s police chief said on Oct 8 that it was not yet clear what lay behind these cases or whether they were linked. One analysis adds an important caution: while agencies have confirmed that attackers use AI to speed up vulnerability scanning and API attacks, there is currently no primary evidence that Japan’s surge in disclosed breaches was caused by Chinese LLMs.
So the careful reading is this: the Korean attacks have been tied to Artex, while the Japanese cases look similar but are not officially confirmed as connected.
Who Is Behind It?
CrowdStrike described the likely attacker as probably a Chinese speaker and financially motivated. Its research linked a single attacker’s IP address to Artex configuration files, session histories for Anthropic’s Claude Code, and a markdown file containing a Chinese-language prompt. South Korean investigators also identified 28 IP addresses tied to the attacks, spread across the US, Japan, Germany and at least ten other countries, which makes attribution difficult.
The key point is that this looks like one person using AI tools, not a large organisation. That is the real shift. AI lets a small actor do work that once needed a whole team.
The Developer Pulls the Plug
On Oct 8, the developer, known as Autumn-27, posted on GitHub that Artex had been abused by some bad actors to launch cyberattacks. The project will no longer be updated and will become closed-source.
Closing the code doesn’t undo the damage, and it doesn’t remove copies already downloaded. Once an open-source tool is released, it effectively stays out in the world.
The Open-Source Dilemma
This case revives a hard debate. Open-source security tools help defenders: researchers can audit them, improve them and use them freely. But as one industry figure pointed out, users of open-source programs can remove the restrictions built into the tool, which makes some kinds of misuse easier.
Closed AI services can refuse harmful requests and ban abusive accounts. Open-source tools running on locally hosted models offer no such checkpoint. That is why regulators, security firms and AI labs are watching this closely.
There is no easy answer. Restricting open-source security tools would hurt defenders as much as attackers, yet leaving powerful automation entirely unguarded invites exactly this kind of misuse.
Why This Matters Beyond Banks
It’s tempting to treat this as a story about large financial institutions. It isn’t only that.
- Smaller organisations are easier targets. Automated tools scan thousands of systems for known weaknesses, and organisations running outdated software get hit first.
- Attacks are getting cheaper. Automation means more attempts, faster, at lower cost.
- Stolen data fuels further crime. Names, phone numbers and income details power follow-up scams long after the original breach.
- Attribution is hard. With attacks routed through many countries, holding anyone accountable takes time.
Practical Steps to Reduce Your Risk
Whether you run a bank or a small company, these basics reduce exposure:
- Patch quickly. Update software, plugins and servers as soon as fixes are released.
- Use multi-factor authentication on every admin, hosting and email account.
- Limit access. Give staff and vendors only the permissions they need.
- Monitor logs and APIs for unusual scanning or login attempts.
- Back up regularly and test that you can restore.
- Minimise and encrypt customer data. Data you never collect can’t be leaked.
- Train your team to spot phishing, since stolen data often fuels follow-up scams.
- Run authorised security audits so you find problems before attackers do.
For individuals: if your bank or a company you deal with reports a breach, change your passwords, watch for suspicious calls or messages that use your real details, and be cautious about unexpected requests for money or verification codes.
What Happens Next
Investigations in both countries are ongoing. South Korea is likely to push for stronger financial-sector security rules, and Japan’s police are still working to establish whether its cases share a common cause. Expect more scrutiny of how AI-enabled security tools are released and who gets to use them.
The Bottom Line
The Artex episode shows AI-assisted hacking is already real, already causing harm, and available to individuals. South Korea has tied it to bank breaches affecting tens of thousands of people. Japan is investigating a parallel wave, with no confirmed link yet. The developer’s decision to close the code is a notable step, but it highlights how hard it is to control powerful dual-use tools once they’re released.
get found in the age of AI.
Frequently Asked Questions
Artex is a Chinese open-source, AI-powered penetration testing tool that uses large language models to automate security testing and hacking tasks.
Not confirmed. Japanese police have said it’s unclear whether the cases are connected or what is behind them. The firmer link so far is to South Korean bank breaches.
South Korea’s Financial Services Commission says more than 68,000 people were affected.
At Shinhan Bank, leaked loan-application information included names, phone numbers and annual income for around 25,000 customers.
CrowdStrike says the threat actor is likely a Chinese speaker and financially motivated, and its research points to a single attacker using AI tools.
The developer, Autumn-27, said bad actors had abused the tool, and the project would stop receiving updates and become closed-source.
Not fully. Copies already downloaded can still be used, and similar tools exist.
No. Open-source tools are widely used for defence. The risk is that any powerful tool can be misused, so the answer is better safeguards and responsible release practices, not avoiding open source altogether.
Yes. Automated scanning targets many systems at once, and those with outdated software are often the easiest to compromise.
Change your passwords, enable multi-factor authentication, watch your accounts for unusual activity, and be wary of calls or messages that use your personal details.
