Tata Electronics, the Indian manufacturer that assembles roughly a third of Apple's iPhones in India, has confirmed a cyberattack affecting part of its IT systems after the extortion group World Leaks began leaking stolen data. The group claims to have taken around 200,000 files, including confidential Apple and Tesla manufacturing and component design documents, internal emails, years of event logs, and copies of employee passports, some belonging to foreign nationals. Researchers say the data has been on the dark web since at least June 10, and a ransom was demanded. World Leaks, a rebrand of the Hunters International group, also claimed breaches at Nike and Dell.
Palo Alto's Unit 42 found a new macOS campaign that uses the ClickFix trick, a fake CAPTCHA or verification page, to get users to paste a command into Terminal. The command quietly downloads a disk image, mounts it without showing it in Finder, finds the app inside, and launches it, installing the Atomic macOS Stealer (AMOS). The malware then shows a fake system password prompt and steals browser credentials and cookies from many Chromium and Firefox-based browsers, cryptocurrency wallet data, Keychain contents, messaging app data, and documents. The single-command approach is stealthier than older campaigns that relied on the victim manually opening a downloaded image.
Security firm AIR showed how easily AI agent skills can be weaponized by building a benign-looking design skill, publishing it to marketplaces, and promoting it with an Instagram ad until it reached roughly 26,000 agents, including some on corporate accounts. Every skill-scanning tool they tested, including offerings from Cisco and Nvidia, marked it safe. The trick is that the skill itself stays clean but tells the agent to fetch instructions from an external page the attacker controls, which passes review while pointing at harmless content and can be swapped for a malicious install script later. Skills load into an agent with the same authority as a user's prompt.
JFrog found malicious npm packages that impersonate PostCSS build tools to drop a multi-stage Windows remote-access trojan on developer machines. One package, postcss-minify-selector-parser, is named to look like the widely used postcss-selector-parser library, which sees over 127 million weekly downloads, and even lists the real package as a dependency to seem plausible during a quick review. Once installed, it writes and runs a PowerShell script that pulls down the trojan. A second cluster of five packages delivers a dropper during npm install, with one server-side component that only serves the payload to victims matching a specific signature. Affected developers should remove the packages and rotate credentials.
Researchers at Calif.io disclosed Squidbleed, a Heartbleed-style memory leak in the widely used Squid web proxy that exposes one user's cleartext HTTP traffic, including passwords, cookies, and session tokens, to anyone else allowed to use the same proxy. The flaw (CVE-2026-47729) is a heap over-read in Squid's decades-old FTP directory parser and is present in the default configuration of every Squid version. To exploit it, an attacker needs proxy access and must point the proxy at an FTP server they control. Only cleartext HTTP and TLS-intercepting setups are exposed; normal HTTPS tunnels are not. A proof-of-concept is public.
Zafran Security disclosed four vulnerabilities, collectively named DifyTap, in Dify, a popular open-source platform for building AI agents and workflows. Two are critical, two need no authentication, and three allow cross-tenant access on Dify's multi-tenant cloud, meaning one customer could quietly read another's private AI conversations and model responses, a covert exfiltration channel. The flaws include an authorization bypass that exposes any application's trace data (CVE-2026-41947), a path traversal into the internal Plugin Daemon API (CVE-2026-41948), and a file-preview authorization bypass (CVE-2026-41949). Most were fixed in Dify 1.14.2, but the path-traversal flaw remains unpatched pending the next release.
FFmpeg has patched PixelSmash, a heap overflow in the MagicYUV video decoder of its libavcodec library that a crafted AVI, MKV, or MOV file can trigger, even during automated thumbnail generation or media scanning. The flaw (CVE-2026-8461) can crash applications or, where address-space randomization is disabled or bypassed, lead to remote code execution; researchers demonstrated full code execution on a Jellyfin media server. Because FFmpeg is embedded almost everywhere video is processed, the bug reaches many self-hosted tools, including Jellyfin, Kodi, Emby, Nextcloud, PhotoPrism, and OBS Studio. The fix shipped in FFmpeg 8.1.2, and several affected projects have updated or added mitigations.
Kaspersky is tracking an active campaign that spreads through WhatsApp by hijacking real accounts and sending their contacts a script file disguised as a business or financial document, with no accompanying message. If a Windows user opens it, the script disables User Account Control protections and silently installs ManageEngine Endpoint Central, a legitimate IT remote-management tool, configured to connect to attacker servers and hand them remote control of the machine. Using trusted contacts and signed, legitimate software helps the attack slip past suspicion and many security tools. The campaign spans several countries, with most confirmed victims in Malaysia, and how the WhatsApp accounts are compromised is still unknown.
Elastic Security Labs detailed OXLOADER, a previously undocumented Windows loader that reaches victims through malicious Google Ads impersonating the Node.js download page and other developer tools. A developer searching for Node.js clicks a sponsored result, lands on a convincing fake site, and runs a script that quietly installs the loader, which then deploys an in-memory infostealer called CastleStealer to harvest credentials and other data. OXLOADER is heavily obfuscated, runs several anti-analysis checks, and skips machines set to Russian or in Russian-aligned regions, pointing to a financially motivated Russian-speaking operator. Google removed the advertiser account, but the technique of buying ads against developer searches remains widespread.
An attacker drained the well-known Ethereum trading bot JaredFromSubway by patiently baiting it into a trap rather than exploiting a software bug. Over several weeks, the attacker deployed 66 fake token contracts and sham liquidity pools mimicking WETH, USDC, and USDT, structured so the bot's automated logic treated them as profitable opportunities and granted token-spending approvals to attacker-controlled contracts. Later trades left those approvals active, and a single transaction then swept the bot's real funds. Security firms estimate the loss near $7.5 million, while the operator claims around $15 million. It is a reminder that standing token approvals in automated systems are dangerous even when the underlying contracts are sound.
TrustStrike Labs