ThreatsDay Bulletin: Worm Code Leaked, AI Agent Phished, Claude Action Patch + 28 New Stories
A weekly security roundup highlights leaked worm code, AI-agent phishing, domain seizures, and a wave of credential-theft tooling.
Intelligence analysis by GPT-5.4 Mini
The bulletin argues that cybercrime is becoming more polished and more industrialized. It pulls together evidence of massive infostealer fallout, a $5,000-a-month RAT, leaked supply-chain attack code, and AI systems that can be tricked into exposing real credentials.
It reads like a report card for internet crime: thieves are selling stronger tools, a secret attack kit was found in the open, and even smart AI helpers can be tricked into giving away keys. The bad guys are running their tricks like a business.
Analysis
What the roundup says
This ThreatsDay bulletin is not a single incident story; it is a snapshot of a security landscape that is getting more industrial. The article opens by arguing that the usual pattern of noisy, sloppy attacks is being replaced by more polished operations, including supply-chain tooling in public repos, credential-stealing RATs sold as a service, and research showing AI agents can be manipulated into leaking real credentials.
The main signals
Several items point to the scale of identity abuse. Flashpoint says more than 11.1 million devices were infected by infostealers last year, feeding a pool of over 3.3 billion stolen credentials, session cookies, cloud tokens, and other identity data across illicit markets. The bulletin also highlights SilabRAT, a malware-as-a-service offering sold for $5,000 a month and focused on credential theft, browser-profile cloning, and cryptocurrency-related artifacts.
The roundup then shifts to state-linked activity and supply-chain risk. CrowdStrike says the North Korean actor Famous Chollima accounted for 47% of state-sponsored hands-on-keyboard operations against the tech sector between April 2025 and March 2026. Separately, the U.S. Department of Justice said it seized 13 domains tied to alleged Chinese intelligence collection efforts that used fake consulting companies and job offers to pressure targets for sensitive information.
Supply chain and platform abuse
One of the most serious items is the Miasma leak. SafeDep says the code was briefly published through compromised developer accounts and describes it as a full supply-chain attack toolkit. According to the article, it can target package registries, GitHub repositories and GitHub Actions, AI coding tool configuration, and SSH-based lateral movement. The bulletin also covers a new RAT for Windows and macOS, a download-pumping trick that inflates npm popularity, an Exchange spoofing weakness, and other recent malware and phishing campaigns.
The common thread is not a single bug but a mature abuse model: stolen identities, trusted platforms, and automation are being combined into repeatable attack systems.
Key points
- Flashpoint says infostealers infected more than 11.1 million devices and fed 3.3 billion stolen identity records into underground markets.
- SilabRAT is being sold as malware-as-a-service for $5,000 a month and is built around credential theft and browser-profile cloning.
- CrowdStrike says North Korea-linked Famous Chollima drove 47% of state-sponsored hands-on operations against the tech sector in the period studied.
- The U.S. DOJ seized 13 domains tied to alleged Chinese intelligence collection using fake consulting and recruiting fronts.
- SafeDep says the Miasma supply-chain attack toolkit leaked briefly and can target package registries, GitHub Actions, and AI coding tools.
If defenders act on these findings, they can hunt for known tooling, tighten account security, and block more stolen-logins abuse before it spreads. Public exposure of the toolkit and domain seizures may also make some campaigns harder to run quietly.
If these trends keep scaling, attackers will keep getting cheaper, easier access to stolen identities and supply-chain compromise. The article suggests the ecosystem is already industrialized, so one weak account, reused token, or risky configuration can still open many doors.
