A phishing operator left a Python web server exposed with directory listing on, letting researchers at Lexfo lift the entire toolkit and pivot to two more operators running three Microsoft 365 phishing campaigns. Each used a custom fork of the open-source Evilginx proxy to steal live session cookies and tokens, defeating multi-factor authentication by relaying the real login. The largest campaign ran over a year, logging 218 captured accounts, mostly corporate mailboxes, and a token file left in a public repository held 97 live Microsoft tokens. The report stresses two different MFA-bypass methods appeared, one proxying the real login and one abusing the device-code flow, which require different defenses to stop.
Nihon Kotsu, Japan's largest taxi operator, shut down internal systems after a cyberattack, disrupting parts of its operations while it investigates and recovers. Precautionary shutdowns like this are a common response to intrusions such as ransomware, aimed at containing an attacker and limiting damage while forensic work proceeds, though they also interrupt normal business. The incident fits a broader pattern of attacks against transportation and logistics providers, whose real-time operations and large customer bases make downtime costly and pressure on victims high. Full details, including whether customer data was accessed, were not yet available, but the disruption underscores how operational technology and business systems in transport are increasingly targeted.
Researchers at the ASSET Research Group demonstrated Ghostcommit, an attack that hides malicious instructions inside a PNG image so AI code reviewers miss them entirely. A harmless-looking conventions file points the coding agent to the image, whose rendered text tells the agent to read the repository's .env file and encode its secrets as a list of numbers written into the code. Because tools like CodeRabbit and Bugbot skip image files by default, the pull request passes review clean; the trap springs later when a developer asks the agent for an unrelated task. The outcome depended on the tool wrapping the model: some agents leaked secrets, while Claude Code refused.
Attackers used a stolen publishing credential to push malicious versions of the jscrambler npm package that install a Rust infostealer as soon as the package is added to a project. Socket ties five bad versions to the same actor, published over about three hours; the malware ships as native binaries for Windows, macOS, and Linux and sweeps developer machines for cloud credentials, crypto wallets, and password-manager vaults. Because jscrambler is a build-time tool used in development and CI, the stealer runs where deploy tokens, cloud keys, and source code live. Partway through, the dropper moved into the package's main code, so installing with scripts disabled does not stop it.
Group-IB detailed a new version of the RedHook Android banking trojan that gains shell-level control of a phone by abusing the operating system's own wireless debugging feature, with no computer or root needed. Once a user is tricked into granting Accessibility permissions, the malware silently taps through Settings to enable Developer Options and Wireless Debugging, then pairs over the loopback interface to run privileged commands. From there it can silently install or remove apps, change secure settings, log keystrokes, stream the screen, and steal credentials, while an aggressive persistence stack keeps it running through reboots. It spreads through social engineering, with attackers impersonating government or bank staff.
Sysdig detailed a financially motivated operator who chained two Langflow vulnerabilities to raid the AI orchestration platform for credentials, part of why CISA just added one of them to its exploited-vulnerabilities catalog. Using a cross-tenant access flaw (CVE-2026-55255), the attacker ran flows belonging to other tenants to steal their large language model provider keys and AWS keys, while a separate unauthenticated code-execution flaw (CVE-2026-33017) took over the host itself. The activity, observed in late June against an internet-exposed Langflow instance, followed a methodical pattern of reconnaissance, flow enumeration, and credential theft. It underscores that AI orchestration platforms concentrate valuable secrets and are increasingly a direct target once left exposed online.
Researchers at Socket uncovered a network of 222 GitHub repositories, tracked as Operation Muck and Load, built to spread Windows malware through a fake Go module posing as a DNS and subdomain scanner. Running the module quietly launches PowerShell that pulls an encrypted payload from attacker infrastructure and deploys infostealers, remote access trojans, and cryptominers. To look trustworthy, the operators used an automated workflow that force-pushes fake commits every minute, making repositories appear actively maintained and generating hundreds of package versions. More than 700 of the module's 1,200-plus versions were malicious. The lures cluster around crypto wallets, Telegram bots, and game cheats, aimed at people likely to run untrusted code.
A cybercrime group left its own server exposed on the internet for weeks, revealing a mass website-hacking operation that researchers at SOCRadar now track as WP-SHELLSTORM. The crew breaks into sites at scale, plants hidden backdoors, and resells the access to other criminals. Rather than using zero-days, it automated attacks against 27 known vulnerabilities in outdated WordPress and Joomla plugins; a single flaw in the Breeze caching plugin accounted for over 17,000 backdoored sites. Target lists named more than 1.4 million domains, with roughly 25,000 confirmed compromised. The same server also revealed an earlier campaign that stole cloud credentials and database passwords from exposed configuration servers.
Attackers compromised a legitimate maintainer's GitHub account for the Injective blockchain SDK and used it to push a malicious version of the widely used @injectivelabs/sdk-ts npm package, which has around 50,000 weekly downloads. The tainted code, disguised as usage telemetry, hooked the SDK's wallet key-generation functions to capture private keys and seed phrases, then sent them to a server made to look like legitimate Injective infrastructure. Trusted-publishing automation spread the malicious release across 18 packages within minutes, though it was live under an hour before being pulled. Because the theft can reach apps that used the SDK only indirectly, any wallet keys handled by affected versions should be treated as compromised.
A new phishing-as-a-service platform called Forg365 is built to steal Microsoft 365 accounts, combining adversary-in-the-middle and device-code phishing with AI-generated lures created directly in its control panel. Researchers at ZeroBEC found the panel lets operators build campaigns, configure malicious OAuth apps, generate and refine phishing emails with AI, and monitor compromised mailboxes for keywords, all in one place. It also ships a browser extension that silently refreshes session cookies through an OAuth flow, giving attackers ongoing access without re-authenticating. The operators deliver lures posing as business documents, using legitimate email-sending infrastructure to slip past filters. The researchers note AI is lowering the cost of both writing phishing content and building phishing platforms.