Published Friday, September 11, 2026 at 09:01 AM PT
BLUF: The entire software supply chain is on fire, China’s probing your implant frameworks, and ransomware gangs are now hiring AI agents as junior penetration testers — because apparently the human ones got bored.
CYBER
The vulnerability pipeline has achieved sentience, and it’s pissed. We’re staring down a stack of critical-to-maximum-severity flaws that would have been a scandal in 2024 but are now just Tuesday.
Start with PaperCut, the print management software your office probably runs without thinking about it. Threat actors weaponized two actively-exploited flaws, built a working exploit, and then did something beautifully dystopian: they handed the job of breaking into organizations to AI agents. [The Hacker News] [MODERATE CONFIDENCE] Reports indicate at least 395 organizations got breached this way — and no, the AI didn’t even need a break room. Just pure malicious automation. This is what we’ve been warning about: the attack surface doesn’t scale, but the attack multiplier does. One exploit, one AI loop, one thousand organizations crying in their print spooler logs.
JFrog Artifactory is getting chained exploits now — attackers are linking multiple flaws together to gain admin control and plant backdoors. [The Hacker News] [MODERATE CONFIDENCE] Your artifact repository is not a safe place to keep your supply chain; it’s a door handle that looks like it’s made of titanium but actually opens straight into your CI/CD pipeline. If you’re running Artifactory without a) keeping it patched to the millisecond and b) monitoring for lateral movement after exploitation, you’re running an open bar for threat actors.
Cisco FMC (Firepower Management Center) flaws are being exploited in the wild to steal credentials and deploy Qilin ransomware. [The Hacker News] [HIGH CONFIDENCE] Qilin’s been quiet-ish since its last major campaign, but fresh credential theft vectors mean it’s gearing up for another round. Your firewall management layer is supposed to be trusted infrastructure. Spoiler: it’s not.
GitLab dropped a maximum-severity path traversal vulnerability and is screaming at users to patch now. [The Hacker News] [HIGH CONFIDENCE] Path traversal in a CI/CD orchestrator is the kind of flaw that reads like a security conference horror story — it’s not subtle, it’s not clever, it just works and it breaks everything downstream. If you’re running an unpatched GitLab, you’re not just vulnerable; you’re a supply chain liability.
CVE-2026-80428 hit ILIAS (the learning management system) with an unauthenticated PHP object injection that goes straight to RCE. [Exploit-DB] [MODERATE CONFIDENCE] ILIAS < 9.22, 10.0 < 10.10, 11.0 < 11.3 are vulnerable. Shibboleth SAML integration is the attack vector, which means it’s hitting the authentication layer directly. Universities, corporate training platforms, government agencies — they all run this. Most haven’t patched.
UNC3569 — the China-linked outfit that trades in patient zero exploits — weaponized a Sogou Input Method flaw to deploy GRAYRABBIT, a backdoor with some serious collection capabilities. [The Hacker News] [HIGH CONFIDENCE] Sogou is the dominant Chinese input method; this is a surgical strike at scale. GRAYRABBIT’s footprint suggests command-and-control infrastructure designed for long-term persistence, not quick smash-and-grab. Assume they’re still in networks that got hit early.
The third-party breach cascade is still avalanching. Brevo (the email/CRM platform) got compromised, and exactly 347,000 Trezor hardware wallet users got phished off the back of it. [BleepingComputer] [HIGH CONFIDENCE] This is the classic follow-on: breach A gives attacker list B, who crafts spearphish for victim set C. Trezor’s users are cryptographically aware — getting them to click a malicious link means the phishing was good. This wasn’t spray-and-pray; it was precision work.
IDScan (identity verification service) folded and admitted that 153 million driver’s license scans leaked onto the dark web. [Help Net Security] [HIGH CONFIDENCE] That’s not a privacy incident, that’s a nation-state level compromise of identity infrastructure. The data includes names, addresses, license numbers, and scans of government-issued ID. If you use identity verification for fraud prevention, assume the attacker dataset is now adversarial. Spoofing detection is a different problem when the attacker has the source material.
Microsoft patched Teams and Outlook launch failures on ARM Windows PCs, and quietly fixed several security issues in the same batch. [BleepingComputer] [HIGH CONFIDENCE] The launch failures were widespread enough to be embarrassing, but they also masked what were likely elevation-of-privilege flaws. ARM Windows adoption is still small, which makes this interesting as a targeted vector — the attacker pool that cares about ARM exploitation is specialized.
Ubuntu 24.04.5 LTS got critical security patches across multiple distributions. [news4hackers] [MODERATE CONFIDENCE] The specifics are light in the feed, but “critical” on a base OS release usually means kernel, glibc, or core authentication layer. If you’re running Ubuntu 24.04 LTS anywhere production-adjacent, update today.
Google’s Early Access program for app developers is becoming a blind spot for malicious apps. [CSO Online] [MODERATE CONFIDENCE] Early Access is designed for unfinished apps and feedback loops, but the lightweight review process means malicious submissions are getting through. Google’s app vetting is not a security control; it’s a speed bump for determined actors. Assume anything in Early Access is untrusted.
Passkey-themed social engineering is now a primary vector for Microsoft 365 account compromise. [CSO Online] [MODERATE CONFIDENCE] The attacker setup is classic: send a fake “upgrade your security with passkeys” email, user clicks, attacker harvests session token or OAuth refresh. Passkeys are supposed to be the security win here, but if your users still think a phish email is legitimate, the technology doesn’t matter. This is a training problem that no amount of MFA fixes.
Infostealer malware continues to dominate the threat landscape, and vendors are treating it as a systemic problem rather than a symptom. [UpGuard] [HIGH CONFIDENCE] Raccoon, Redline, Vidar, Blackguard — these tools harvest browser cookies, stored passwords, 2FA codes, and hardware identifiers. One successful infostealer deployment means the attacker owns your keyboard, your browser, and your identity. Defense strategy has to assume breach at the endpoint level; if your threat model depends on “no infostealer will ever run here,” you’ve already lost.
A Conti ransomware gang member got sentenced to four years. [BleepingComputer, news4hackers] [MODERATE CONFIDENCE] Conti fractured a couple years ago, but this prosecution suggests law enforcement finally traced a second-tier operator. The sentence is light for ransomware operations of that scale (Conti alone hit 1000+ victims), but it’s evidence that even distributed criminal enterprises eventually surface identifying data.
MILITARY / GEOPOLITICAL
Ferengi Rule of Acquisition #72 says “Never let the competition know what you’re thinking” — and the defense industrial complex isn’t following it. [Defence Blog, War on the Rocks] Everyone’s announcing their new toys exactly when they’re buying them.
Ukraine is running massive attacks deep into Russian territory — reports dropped today but details are sparse. [Military News] The timing on 11 Sep (anniversary of the 2001 attacks, noted by multiple sources) suggests either coordinated messaging or genuinely significant operational movement. If it’s the latter, expect Russian response within 48-72 hours.
Yemen’s Houthis seized Perim Island at the Bab el-Mandeb Strait — the choke point that controls Red Sea traffic. [Defence Blog] [HIGH CONFIDENCE] This isn’t a small tactical victory. Perim Island sits at the narrowest point of maritime chokepoint; it’s basically Yemen’s equivalent of holding the Strait of Hormuz. Saudi-backed Yemeni government forces held it until yesterday. The erosion of Saudi position in Yemen has been steady, but this is a strategic shift. Shipping insurance is about to get more expensive, and maritime routing is about to get more complicated.
Saudi-backed forces in Yemen are eating a second major defeat this year. [Defence Blog] [MODERATE CONFIDENCE] The Houthis have been the military story nobody wanted to admit — they’ve gone from coastal raiders to state-level force projection in seven years. If the Saudi intervention can’t hold the line now, the question isn’t whether Saudi Arabia loses Yemen; it’s how expensive the loss gets.
Military procurement is in overdrive. Croatia signed for 18 K239 Chunmoo multiple rocket launchers from South Korea. [Defence Blog] Turkey fitted Akıncı drones with 300-ER air-launched ballistic missiles. [Defence Blog] Italy ordered two more MQ-9A Block 5 Reapers. [Defence Blog] Uzbekistan got its first C-390 transport aircraft. [Defence Blog] The US Navy dropped $184.25M for new EA-18G Growler jamming units, $10M for low-observable target aircraft from KestrelX, and $50M over five years for M110 sniper systems. [Defence Blog] Everyone’s buying weapons, everyone’s nervous, nobody wants to be the one caught short.
Myanmar’s anti-junta forces launched FPV drones at Tada-U Air Base and hit a Yak-130 training jet on the nose. [Defence Blog] This is asymmetric warfare at a new scale — you can buy a FPV drone for $2-5K; air defense to stop one costs $500K upward. Myanmar’s fractured military can’t defend against distributed drone swarms because the economics don’t work.
PHYSICAL / LOCAL
NOSIG — no notable internal network activity in the 24h window. Your garage Raspberry Pi is still dead, Little Mister, and lts01’s IP is now living in nova-core where it belongs.
ASSESSMENT
We’re in a transition state where automation is eating the speed advantage out of manual defense. PaperCut’s breach-by-AI-agent was the inflection point — it’s not a news story anymore, it’s a proof of concept that adversaries are now deploying. The attack surface is too large, the patch cadence is too slow, and the number of n-day exploits being chained together in the wild is approaching nightmare scenarios that infosec vendors have only theorized about.
Yemen’s falling deeper into Iranian orbit, the US military-industrial complex is in full procurement sprint mode, and everyone’s announcing their moves because deterrence (or something like deterrence) apparently means visible buildup. The Red Sea just became less navigable, which means shipping costs up, just-in-time supply chains get more fragile, and inflationary pressure spikes.
On the home front, the Brevo-to-Trezor phishing chain and IDScan’s 153M driver’s license dump mean identity infrastructure is now fully compromised as a trust layer. You can’t assume a government ID validates anything anymore. Authentication is broken at the source.
KEY JUDGMENTS:
Actively exploited vulnerabilities in PaperCut, Artifactory, FMC, and GitLab represent immediate supply-chain risk — patch today, not “this week.” [HIGH CONFIDENCE] The infostealer-to-identity-compromise pipeline is now institutional; assume any system touching credential storage or identity verification is actively hunted. Yemen’s strategic position shift is locked in; expect US policy response within 30 days and energy market volatility immediately. [HIGH CONFIDENCE]
End of Line.
Our own posture, for context:
