In partnership with
Scale AI support on AWS, see how July 9
Customer expectations keep rising. Support budgets don't. On July 9, Fin and AWS are hosting a live executive session on how leading enterprises close that gap: scaling AI-powered support while simplifying how they buy it.
You'll see how to resolve an average 76% of conversations with Fin on AWS enterprise-grade infrastructure, procure through AWS Marketplace to put committed cloud spend to work, and turn the Fin and AWS collaboration into lower support costs. Register for the live session to see how.
This week, the threat signal clustered around six recent developments: North Korean actors flooding open-source ecosystems with malicious packages and extensions, a Linux kernel flaw enabling root escalation, an Opera GX browser flaw allowing silent mod installation and data theft, Sysdig’s documentation of agentic ransomware behavior, law-enforcement and Google disruption of NetNut proxy infrastructure, and a reported government extortion payment tied to stolen files.
This is not a “single bad CVE” week.
It is a trust-surface week: Developer tooling, browsers, Linux kernels, AI workflows, residential proxies, and government data all showed up on the board.
Trend (Macro) | Likelihood | Direction | Signal for the Week |
|---|---|---|---|
Open-source supply-chain poisoning | 84% | 🔺 Rising | PolinRider published 108 malicious packages and extensions across npm, Packagist, Go, and Chrome. |
Linux local privilege escalation | 78% | 🔺 Rising | Bad Epoll gives unprivileged users a path to root on affected Linux and Android-adjacent systems. |
Browser-side data theft | 72% | 🔺 Rising | Opera GX showed how browser customization features can become silent data-exfiltration paths. |
Agentic ransomware / AI-run intrusion chains | 70% | 🔺 Rising | JADEPUFFER demonstrated automated intrusion, credential theft, encryption, and extortion behavior. |
Residential proxy abuse and disruption | 68% | ➡ Stable | NetNut disruption shows proxy networks remain key attacker infrastructure for hiding activity. |
Data-theft extortion without classic ransomware | 66% | 🔺 Rising | Kairos-style extortion shows stolen-file leverage remains enough to force payment decisions. |
North Korean PolinRider campaign floods developer ecosystems — Threat actors linked to Contagious Interview published 108 malicious packages and browser extensions across npm, Packagist, Go, and Google Chrome, making PolinRider a developer-workstation and CI/CD credential-theft concern, not just “another bad package.”
Bad Epoll Linux kernel flaw enables root escalation — The newly surfaced CVE-2026-46242 use-after-free race can let a local user climb to root on affected Linux systems, making Bad Epoll a priority for shared servers, container nodes, developer endpoints, and Android-adjacent fleets.
Opera GX flaw allowed silent mod installation and data theft — Opera patched a vulnerability that could let a malicious site silently install a GX mod and use CSS-based techniques to exfiltrate page data, including a signed-in Gmail address in proof-of-concept testing, making GX mods a browser-governance issue for users and teams that allow nonstandard browsers.
JADEPUFFER shows agentic ransomware is no longer theoretical — Sysdig documented what it assesses as the first known agentic ransomware operation, where an LLM autonomously used a Langflow RCE path, harvested credentials, pivoted, encrypted data, and attempted extortion, making JADEPUFFER a wake-up call for exposed AI workflow platforms.
Google and FBI disrupt NetNut residential proxy infrastructure — Google, the FBI, and partners disrupted NetNut-linked infrastructure used to mask malicious traffic, with Google saying the action reduced the available proxy pool by millions of devices, making NetNut a useful reminder that attacker anonymity depends on infrastructure that can be degraded.
U.S. government entity reportedly paid Kairos in data-theft extortion case — A July 04 case study reported a roughly $1 million payment to prevent stolen files from being leaked, making Kairos a reminder that encryption is no longer required for extortion if stolen data creates enough pressure.
200+ Proven Ways to Make Money With AI in 2026
The next wave of millionaires will be people who figured out how to make AI work for them.
The window to get ahead is still open. But not for long.
Here are 200+ proven ways to make money with AI in 2026.
Sign up for Superhuman AI, the free daily newsletter read by 1M+ professionals, and get instant access to all 200+ ways to profit from AI this year.
Developer trust is still cheap to abuse. Malicious packages and extensions are scaling because developers install tools faster than security teams can review them.
Local privilege escalation matters in cloud-heavy environments. “Local user required” is not comforting when attackers already have code execution through containers, CI runners, or compromised developer systems.
Browsers are becoming data planes. Extensions, mods, and customization systems can become quiet collection mechanisms when governance is loose.
AI workflow platforms are becoming attacker infrastructure. Langflow-style tools often hold API keys, cloud credentials, and automation paths that turn one exposed app into a pivot platform.
Proxy infrastructure remains a criminal force multiplier. Residential proxies help attackers bypass geolocation, reputation filters, and login defenses.
Extortion is shifting from encryption to leverage. If stolen files are sensitive enough, the attacker does not need ransomware. They just need pressure.
Supply-chain containment: Search for PolinRider-linked packages and extensions, block risky package sources, rotate developer tokens, and review CI/CD secrets exposed to developer endpoints.
Linux patch prioritization: Prioritize Bad Epoll remediation on shared Linux hosts, container nodes, build runners, developer systems, and Android-adjacent assets; hunt for unexpected root-owned processes and kernel crash artifacts.
Browser governance: Confirm Opera GX version 130.0.5847.89 or later where present, restrict unsupported browsers in enterprise environments, and review extension/mod policies.
AI workflow exposure review: Identify exposed Langflow or similar AI workflow platforms, patch known RCE paths, rotate stored API keys, and restrict admin interfaces to trusted networks.
Proxy-aware detection: Enrich authentication telemetry with residential proxy intelligence, watch for rotating consumer IPs, and tune password-spray detection for low-and-slow distributed attempts.
Data-extortion readiness: Confirm what data is externally reachable, test legal/comms decision paths, and ensure leadership knows the difference between “encrypted systems” and “stolen files with public-pressure leverage.”
The good news: Google, the FBI, and industry partners just showed that criminal proxy infrastructure can be degraded at scale. NetNut-linked operations lost millions of available devices, which means attacker anonymity got more expensive and less reliable.
That matters.
Criminal operations depend on infrastructure, and infrastructure leaves fingerprints.
Keep improving telemetry, keep sharing indicators, and keep making the bad guys rebuild their toys.
This weekend’s lesson: Attackers are targeting the things teams trust by default: packages, browsers, kernels, AI workflows, and “normal-looking” residential traffic. Trust is fine. Unchecked trust is just an incident waiting for a timestamp.
J.W.
(P.S. Forward to your CISO / Add to Board Briefing.)
See what enterprise-ready AI support looks like
How are leading teams getting AI support to work? We're breaking down the playbook, live July 9.
