🚨 ACTIVE EXPLOITATION
- North Korean WaterPlum hackers infected 30,000 devices, stole $10.7M in crypto
A joint US, Japanese, Australian, and German advisory says North Korean WaterPlum hackers compromised 30,000 devices and moved $10.7 million in stolen crypto to North Korea.- WaterPlum targeted job seekers worldwide, impersonating AI, cryptocurrency, and NFT companies on recruiting and freelance platforms in the 'Contagious Interview' campaign.
- At least 30,000 devices in over 100 countries were infected from December 2025 through July 2026, with funds or credentials stolen from over 7,000 cryptocurrency wallets.
- Malware families include BeaverTail (npm), InvisibleFerret (Python backdoor), OtterCookie, OtterCandy, and StoatWaffle (Node.js malware in Visual Studio Code projects).
- Attackers steal browser credentials, clipboard contents, keystrokes, crypto private keys and seed phrases, and may pivot to employers' networks for IP theft and espionage.
- WaterPlum overlaps with North Korea's fraudulent IT worker scheme under the 313 General Bureau, using AI face-swapping in interviews and stolen identity documents to obtain jobs.
💥 BREACHES & INCIDENTS
UPDATE: IDScan confirms hackers stole 13-15 million driver's licenses in months-long breach
IDScan confirmed hackers stole at least 13-15 million driver's licenses and passports during a five-month breach and appear to be extorting the company for ransom.- IDScan, a Louisiana-based identity verification provider used by entertainment venues, car rentals, and dispensaries, confirmed the breach in a September 17 disclosure to state authorities.
- Stolen data includes full names combined with dates of birth and driver's license, passport, and other government-issued ID numbers, affecting an estimated 13-15 million people.
- Hackers had access to IDScan's systems from April 1 to September 2, 2026; the company learned of the breach around September 1, including via information from the FBI.
- Hackers claimed over 150 million U.S. and Canadian driver's licenses scraped during an alleged year-long breach, exposed via a dark web search site that disappeared after Krebs reported it.
- IDScan restricted access to its APIs, including for client accounts, or decommissioned them, suggesting APIs may have been the theft vector; the company implied hackers are demanding payment for full access to the stolen data.
Sources 📄 Source: documentcloud.org · 📎 Coverage: this.weekinsecurity.com · 👁 via @zackwhittaker@mastodon.social (+1)ShinyHunters hacks Clop leak site, threatens to extort ransomware gang
ShinyHunters breached and defaced the Clop ransomware gang's data leak site and claims to have stolen its server data and onion private keys.- ShinyHunters breached the Clop ransomware operation's Tor data leak site, defacing it with Umbreon ASCII art and a message mocking the gang.
- The attackers exploited what they claim is an unauthenticated file upload vulnerability in Grav CMS, first uploading a taunting text file, then fully defacing the site.
- ShinyHunters claims it gained full server access and stole source code, Grav CMS plugins, /var/log files, and the private keys for Clop's onion service, which would let it host the same onion URL.
- The attack is retaliation in an ongoing feud dating to Clop's 2025 Oracle E-Business Suite extortion campaign, which used zero-day CVE-2025-61882; ShinyHunters says a Clop representative threatened its members.
- ShinyHunters says it plans to extort Clop, giving the gang 72 hours to make contact via its own leak site; BleepingComputer confirmed the defacement but not the data theft claims.
🔓 CVEs & KEV
CVE-2026-93985 — Openpanel-dev openpanel — CVSS 9.4 — OpenPanel js-runtime JavaScript Template Sandbox Escape RCEOpenPanel js-runti...
CVE-2026-93984 — Openpanel-dev openpanel — CVSS 6.9 — OpenPanel API Authentication Bypass via Unverified Client SecretOpenPanel tra...
🕵️ RESEARCH & DEEP DIVES
- Researchers Used Claude Opus 5 to Chain Flaws and Take Over OpenAI Staff Accounts
Hacktron researchers used Claude Opus 5 to chain a libheif forum bug with an OpenAI SSO weakness and take over several OpenAI employees' ChatGPT and Codex accounts.- Affected: OpenAI staff ChatGPT/Codex accounts and self-hosted Discourse forums running Debian 12 images with unpatched libheif 1.19.7
- Impact: account takeover via shared 'Sign in with OpenAI' SSO, reaching an internal OpenAI code repository; wider access to GitHub, Slack, and email was possible but not used
- Attack chain: crafted HEIC/HEIF image triggers libheif out-of-bounds read (CVE-2026-32882, 8.8/10) defeating ASLR, escalated with Claude Opus 5's help to code execution on the forum server, then pivoted through OpenAI's login system
- Claude Opus 5 produced a working exploit within hours after Opus 4.8 failed; safeguards were bypassed by framing the target as a CTF practice server
- OpenAI fixed the flaw ~14 hours after the report and paid a $6,500 bounty on September 1; libheif fixed in 1.22.0 (May 2026), latest release 1.23.4
📋 ADVISORIES
- 📄 Source for Metacurity weekly roundup: AI safety vindicated, AI warfare, catfishing, and Flock surveillance — theverge.com