Topic
Use of general-purpose AI coding agents and LLMs by threat actors across reconnaissance, exploitation, data triage and exfiltration.
Zscaler ThreatLabz says a threat actor used the ARTEX AI pentest framework to steal personal data on about 67,000 people from South Korean financial firms. Entry was through peripheral web apps. Zscaler rates its attribution to a Chinese-speaking, financially motivated actor at moderate confidence.
Reality
ARTEX's developer made the open-source agent closed-source on October 8 after CrowdStrike linked it to a China-based 26-year-old who breached South Korean banks. The campaign exposed roughly 68,000 customer records, Cryptobriefing reported.
Perspective Coverage
5 publishersReality
Hacktron chained an image-library bug in OpenAI's Discourse forum to a single sign-on misconfiguration, reached employee ChatGPT and Codex accounts inside 72 hours, and the repair OpenAI describes is a narrowing of token permissions.
Perspective Coverage
9 publishersReality
CrowdStrike tied AI-assisted South Korean bank attacks, run on an open-source pentest tool and DeepSeek, to a possible 26-year-old suspect in Guangdong. The reported breaches involved systems banks run for brokers and staff, so the exposure to manage sits at the bank's edge, whoever the attacker proves to be.
Perspective Coverage
6 publishersReality
South Korea's FSC paused a plan to open bank network-separation exemptions to 75 firms after ARTEX-assisted hacks breached seven financial firms. The pause holds back at most 15 exemptions while regulators work out whether the attacker got in through the rule they were loosening or around it.
Perspective Coverage
8 publishersReality
Oasis Security says attackers' servers held about 960,000 Yoido Full Gospel Church member records with resident registration numbers. The church is still verifying, but the cache shows how much identity and donation data sits behind one administrator account at a large congregation.
Perspective Coverage
3 publishersReality
President Lee Jae-myung ordered a full probe of hacks that exposed data at four Korean banks, including about 25,000 Shinhan Bank customers. The response runs through every financial industry association before the probe has shown whether the banks share a weak point.
Perspective Coverage
5 publishersReality
Shinhan Bank exposed data on about 25,000 customers through a loan-broker inquiry platform kept apart from its core banking systems. Yonhap reports AI may have automated the credential stuffing, so the side systems that touch customer data are where lenders' security spending now gets tested.
Perspective Coverage
4 publishersReality
Azazel, a Russian-speaking affiliate of The Gentlemen RaaS crew, ran his own leak site to extort over two dozen victims and kept the proceeds, CloudSEK found. His way in was secrets left in old GitLab commits.
Reality
Korean regulators sent security orders to about 500 financial firms after the same attacker IP turned up in breaches at seven lenders. Almost every firm doing the emergency reviews has no reported breach, so the cost is shared across the sector.
Reality
South Korea's Financial Services Commission ordered banks and card firms to check every externally reachable IT system after data leaks at four banks. The regulator will collect the returns and use them to build new cybersecurity measures for the financial sector.
Reality
DIVD said attackers chained two Zammad zero-days on its own internet-facing server and went from a hijacked session to root in seconds. It assessed that an AI agent was involved and says upgrading to version 7 is not a complete fix for both flaws.
Reality
Gambit Security says the ransomware crew used a commercial coding agent for hands-on post-compromise work between 8 April and 21 May, alongside a new Linux encryptor that force-kills running guests before it touches ESXi datastores.
Perspective Coverage
4 publishersReality
Attackers using AI can fix and retest a failed cloud privilege escalation in minutes instead of hours, according to an analysis in The Hacker News. The cost it puts on defenders is the hours spent rebuilding each alert's context before anyone acts.
Reality
Sysdig's threat research team watched one operator work a marimo notebook host for nine hours with hand-written scripts, and the eight-second jump to a bastion host at the end ran on tooling already staged on disk.
Perspective Coverage
3 publishersReality
Spain's AEPD published the account but has not yet verified it. In it, the agent scanned for flaws, logged in, altered personal data and opened invoices. The agency's own position is that AI adds speed and scale, not new threats.
Perspective Coverage
3 publishersReality
The image parsing bug had been fixed upstream about a year earlier, and the decision not to ship that fix belonged to Discourse. OpenAI's forum shared single sign-on with internal systems, so a forum account became GitHub access.
Perspective Coverage
9 publishersReality
Gambit Security says one operator used three AI agent tools to steal over 600,000 cards, spending an estimated $12,000 to $18,000. One retailer's scheduled job reinstalled a skimmer after a deploy removed it, so recovery tests have to cover every layer the agents touched.
Reality
Three cybersecurity specialists told The Verge the attackers they worry about are humans holding generative AI, aimed at infrastructure where the average US nuclear reactor is about 44 years old and some equipment vendors no longer exist.
Reality
The Spanish agency describes a third party using an agent to chain a successful login, vulnerability discovery, data modification and invoice access, and it published that account while its own investigation is still open.
Reality
Security · September 14, 2026 · One report1 publisher
Invest · September 7, 2026 · Confirmed3 publishers
Science · September 5, 2026 · Confirmed2 publishers