Researchers at VulnCheck found a flaw in the DeepSeek Harness, a tool that runs an AI agent's commands inside an operating-system sandbox so an agent handling untrusted files cannot write outside its workspace. Through an authentication bypass using a spoofed host header, an attacker needing no credentials or API key can call the tool's own web interface to invoke privileged commands with full-access permissions, raise the session's approval policy to unrestricted execution, and read every stored conversation. In effect, the sandbox meant to contain the agent can be switched off from outside. It is a reminder that an AI agent's isolation is only as strong as the authentication protecting its control interface.
Researchers at Sophos and ESET found attackers breaching F5 BIG-IP APM access gateways and installing a stealthy Linux rootkit that ESET calls PoisonedRefresh. Rather than dropping a file on disk, it hides a web shell in memory, hooks into the Apache and PHP components, tampers with SELinux settings, and uses disguised requests to run commands while blending into normal traffic. Crucially, it persists across device upgrades, so patching the appliance alone does not remove it. The intrusions likely began by exploiting a critical remote code execution flaw that F5 had earlier downgraded to a mere denial-of-service issue, which may have led some organizations to deprioritize patching it.
Security researchers built a zero-click worm that hijacks WeChat accounts through an incoming voice call on both iPhone and Android, without the target ever answering. The exploit fires during the ringing phase, before the user declines or picks up, abusing a memory-corruption flaw in WeChat's call-handling code to run commands on the device and take over the account. Because it can spread from a compromised contact to their contacts, it behaves like a worm. Notably, the researchers used AI to find the bug and write the exploit in about two days. Tencent patched it in late August and added a server-side mitigation, and saw no in-the-wild abuse before the fix.
Google's threat-intelligence team reported that a financially motivated attacker used an autonomous, multi-agent AI framework to compromise thousands of third-party credentials in under six hours, a pace that would take a human far longer. It is part of a broader shift in which criminals fold autonomous and coding-focused AI agents into operations, using them to get past defenses, reverse-engineer software, and sift stolen data. Google also found infostealer malware now specifically targeting AI developer configurations and credentials, and actors running open-weight models on compromised machines to sidestep restrictions on commercial services. The takeaway is that AI is collapsing attack timelines from days to hours, and AI access itself is now worth stealing.
Researchers disclosed a post-exploitation toolkit called PEEP that turns Chrome and Edge into backdoors for stealing credentials and running commands on the host. After an attacker already has administrative or code-execution access, PEEP's installer injects a malicious extension disguised as a bookmarks tool directly into browser profiles, forging Chromium's own integrity settings to bypass the Web Store and skip user approval prompts. The planted extension then serves as a covert channel to harvest credentials and execute operating-system commands. Because it abuses a trusted, ever-present browser and hides its extension from the usual checks, it can persist quietly on a compromised machine, a reminder that browsers themselves are a rich post-compromise attack surface.
Researchers at CloudSEK gained access to the control panel of BigBear, a phishing-as-a-service platform that defeated multi-factor authentication at 258 organizations and stole thousands of Microsoft 365 credentials. It uses an adversary-in-the-middle proxy based on Evilginx to sit between victims and Microsoft's real login, capturing passwords, multi-factor codes, and the session cookie, then replaying the cookie to hijack the already-authenticated session. The panel logged over 5,000 stolen records across 40-plus countries and is rented to multiple affiliates who receive stolen data through Telegram bots. Notably, it runs JavaScript that disables the browser's passkey support, forcing victims off phishing-resistant login onto weaker methods it can intercept.
Researchers at Check Point analyzed JSCeal, malware compiled into a hard-to-analyze bytecode format that steals browser cookies and authentication tokens to hijack accounts. By replaying stolen session cookies, an attacker can access a victim's Google account without the password or a second factor, and the malware also grabs saved passwords, autofill data, and OAuth tokens to automate further account access. It additionally targets cryptocurrency wallets and platforms and includes keylogging, screenshots, and messaging-session theft. The compiled format and layered obfuscation push it outside analysts' usual tooling, though Check Point released a deobfuscator. It is a reminder that stolen session cookies quietly defeat passwords and multi-factor authentication alike.
Researchers at Elastic documented four persistent programs tied to the REVSTEALER infostealer that stay on a machine even after the stealer deletes itself. One disables Windows Update services and Microsoft Defender, adds Defender exclusions, and kills update and malware-removal tasks before hiding a cryptocurrency miner inside legitimate Windows processes. The malware also bypasses Chrome's app-bound encryption by launching the browser in a debugger to read the decryption key from memory, and steals session cookies to take over accounts without passwords. It spreads through game-cheat lures on hijacked video channels and pirated or fake application installers. Because these modules outlive the stealer, a confirmed infection warrants reimaging rather than cleanup.
Researchers at GitGuardian found that a recent variant of the self-spreading Shai-Hulud npm worm has expanded its credential theft to scan 469 distinct locations on infected developer machines. The targets now span developer environments, continuous integration and deployment tooling, cloud configuration files, and even the configuration of AI tools. That breadth turns a single compromised package into a wide net for secrets, from cloud and registry credentials to keys held by developer and AI tooling. It reflects how supply-chain worms are industrializing secret collection, treating any credential a developer's machine can reach as fair game once malicious code runs during installation or use.
Symantec reported that threat actors are abusing the legitimate, digitally signed Node.js runtime to run malicious JavaScript while slipping past security tools, in attacks on government, technology, and hospitality targets since February. Because the Node.js executable is a trusted developer tool, defenses rarely flag it, so instead of dropping a malicious program the attackers stage the genuine runtime and keep their harmful logic in interpreted scripts. They gain persistence through a Windows registry startup key and, in one case, pulled command-and-control instructions from the blockchain using a technique called EtherHiding. The activity has been tied to a ClickFix social-engineering entry point and an initial-access broker.