Researchers Expose AI-Driven Bank Attack Tools Left Open by Hackers
Insikt Group uncovered open infrastructure revealing AI-assisted attack preparations against South Korean financial institutions.
Key points
- Exposed files revealed a wider operation involving cryptocurrency theft and Linux attacks beyond the banking campaign.
- Public reporting indicated breaches exposed personal data of at least 68,000 people across seven financial institutions.
- Recorded Future tracked the actor, TAG-217, using AI tools like Claude Code and ARTEX for penetration testing.
Researchers from Recorded Future’s Insikt Group discovered open infrastructure belonging to hackers linked to AI-assisted attacks on South Korean banks. The exposed files provided a detailed view of the threat actor’s tools and preparations, including scripts and configurations used during the campaign.
The investigation, detailed in an October 16 report, traced the operator’s activities back to August 2026. The findings showed how basic security mistakes by the attackers allowed researchers to reconstruct their workflow. The group, tracked as TAG-217, left directories open, revealing evidence of their broader operational scope.
Context and Background
The campaign utilized ARTEX, a Chinese-language open-source framework designed for authorized penetration testing. Attackers repurposed this tool to probe targets without permission. Earlier analysis dated October 7 linked the activity to several AI platforms, including Claude Code, DeepSeek, Grok, and Zhipu AI.
Public analysis noted exposed Claude Code sessions and AI memory files that helped reveal the operator’s workflow. Investigators suspected a Chinese-speaking operator based on language clues in the tools. However, the source material notes that language indicators do not establish nationality or government backing for the group.
Recorded Future assessed that the broader activity likely reflected attempts to generate revenue across multiple targets. The incidents drew attention to systems outside core banking services. The operator tested infrastructure for maintaining remote access and scanned internet-facing systems between August 17 and 22.
Who Is Affected
Public reporting cited in the investigation stated that breaches exposed personal information belonging to at least 68,000 people. These incidents occurred across at least seven financial institutions in South Korea. Different organizations lost different types of data, rather than suffering one identical breach across all entities.
The compromised data included customer records and personal details of an outsourced developer. Reporting on the security investigation described breaches affecting loan-agent and employee-support systems. These supporting services hold valuable personal and credit information, distinct from primary customer banking platforms.
The exposure of these supporting systems highlights why security reviews must cover connected business applications. Core banking platforms are often heavily fortified, but peripheral systems can still contain sensitive data. The varied nature of the data loss indicates a targeted approach to different organizational assets.
What should you do?
Organizations should review their own internet-facing systems for open directories and exposed configuration files. Security teams need to audit connected business applications, such as employee support and loan-agent systems, for vulnerabilities. Monitoring for unusual AI tool usage or unauthorized penetration testing frameworks is also recommended.
What to do and how to stay safe: TAG-217
- Audit internet-facing systems to ensure no directories or configuration files are publicly accessible.
- Review security controls on peripheral business applications, not just core banking or primary services.
- Monitor network logs for signs of unauthorized penetration testing tools or remote access attempts.
- Verify that AI development environments are isolated from production infrastructure to prevent data leakage.
Step-by-step guide: Social Engineering Warning Signs and How to Respond
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Which AI tools were used in the attacks on South Korean banks?
Investigators linked the activity to Claude Code, DeepSeek, Grok, and Zhipu AI, based on exposed sessions and memory files.
How many people were affected by the data breaches?
Public reporting indicated that personal information belonging to at least 68,000 people was exposed across seven financial institutions.
Who is the threat actor behind these attacks?
Recorded Future tracks the actor as TAG-217, suspected to be a Chinese-speaking operator, though nationality is unconfirmed.
