The Hacker News | #1 Trusted Source for Cybersecurity News — Index Page

AI agent security is moving through a familiar maturity curve: adoption, then visibility, and finally, control. But what we've collectively discovered is that enforcing least privilege for AI agents is harder than we ever imagined. This is why there are so many approaches, from prompt filtering to identity-layer access controls. Where we've collectively landed is that understanding the intent of AI agents is essential to securing them. It's not easy, but it's the only path forward. Organizations approach this challenge with different levels of sophistication. For many, the current goal is simply to find the AI agents already operating across the business. That is a necessary first step. AI agents are appearing in SaaS platforms, developer environments, cloud workflows, customer support systems, productivity tools, and internal applications. Some are sanctioned, and others are not. But discovery alone isn't useful. AI agents are not passive; they reason, plan, c...

Someone installed a popular AI assistant on a rented server, switched off the setting that makes it ask permission before running risky commands, and pointed it at Thailand's Ministry of Finance, which runs the country's treasury and tax collection. The agent then worked through the ministry's network on its own, checking hosts for ways to gain root access, hunting through file systems, and crawling a folder of staff personnel records going back to 2012. The operator left the agent's own logs sitting on a web server with directory listing switched on, where threat intelligence firm Hunt.io and researcher Bob Diachenko found them, along with 585 files and 470 MB of attack tooling. The tool is Hermes , an open-source assistant from Nous Research that people install to manage their mail, run chores, and take instructions over Telegram or Slack. It is not a hacking tool, and nothing here is a flaw in it. The mode the operator used, called YOLO, is a documented feat...

The threat actors behind the Golden Chickens malware-as-a-service (MaaS) ecosystem have resurfaced with four new malware families, indicating that the operators are showing no signs of stopping despite extensive public disclosures into their inner workings. The malware families in question are: TinyEgg, ChonkyChicken, a modularized variant of ChonkyChicken, and a modified web browser credential theft utility codenamed ChromEggscalator. Recorded Future's Insikt Group is tracking the group under the moniker TAG-195. TAG-195 is a financially motivated malware-as-a-service (MaaS) developer whose tooling has been previously linked to TAG-127 as an operator and customer. The threat intelligence company said it has also observed TAG-127 deploying TinyEgg via ClickFix-style social engineering campaigns that trick unsuspecting users into manually executing malicious commands. "The four new families indicate an architectural transition and evolution in the TAG-195 MaaS ecosys...

Eight security flaws in NodeBB went public on Wednesday, along with the code to exploit them. Aikido Security rates all eight as high severity and says its AI pentest agents found them in a six-hour review of the forum software's source code. Every version before 4.14.0 is affected. NodeBB has fixed them all, and administrators should be on 4.14.2. The simplest one takes a settings change. A regular forum member could point their homepage setting at the admin address, reload the page, and the admin dashboard opened for them. No password, no exploit code. The forum's own interface blocks that setting, but the block only ran in the browser and could be sidestepped. Most of what a member could then reach was read-only, including the error log and any user list an admin had exported, though they could also swap the site logo. Two more gave an attacker with no account at all access to things meant to be private. One let anyone claim to be any user and read private messages...

Redis shipped seven security releases on July 23 after researchers published authenticated RCE PoCs for stock Redis 6.2.22, 7.4.9, 8.6.4, and 8.8.0. All four chains require RESTORE. The Streams chains also need EVAL and XGROUP; the 8.8.0 chain needs EVAL and the bundled RedisBloom module. Redis says the underlying memory flaws may lead to remote code execution . Redis 6.2.23, 7.2.15, and 7.4.10 fix the Streams shared-NACK use-after-free; Redis 8.2.8, 8.4.5, and 8.6.5 fix both the Streams issue and the RedisBloom and TDigest out-of-bounds writes; Redis 8.8.1 fixes the RedisBloom and TDigest loaders, while the Streams guard was already present in Redis 8.8.0. Two PoC targets, Redis 6.2.22 and 7.4.9, were the May security updates Redis told users to install, but those releases did not include the shared-NACK ownership guard. Upgrade to the fixed release for the deployed branch. Until then, revoke RESTORE from accounts that do not strictly need it and block untrusted network acce...

The Computer Emergency Response Team of Ukraine (CERT-UA) has warned of a new campaign that involves the use of a malicious program that's dressed up as a Notepad++ plugin to compromise Windows systems. The activity has been attributed by the agency to a threat cluster it tracks as UAC-0099 , a Russia-aligned group that has previously observed weaponizing security flaws in WinRAR software to deliver a malware strain called LONEPAGE. Other cyber attacks mounted by the adversary have employed phishing emails as an initial access method to deploy MATCHBOIL, MATCHWOK, and DRAGSTARE. It's known to be active since at least mid-2022. The latest set of attacks begins, observed earlier this summer, with a phishing email containing an image attachment, which, when clicked, opens a URL that's concealed using a link shortener from where the request is sent to a file-sharing service like EasySend[.]co to retrieve a ZIP archive. The ZIP file contains a Visual Basic Script (VBSc...

A Russian state-supported espionage group spent months reading Western mailboxes through a then-unknown flaw in Zimbra's webmail client. The payload goes after the last 90 days of email, the organization's entire email directory, the password saved in the browser and the codes kept for two-factor recovery. Opening the message was enough to start it. The NSA , CISA and partner agencies published a joint advisory on the campaign Thursday, alongside research from Palo Alto Networks' Unit 42 and Proofpoint. The advisory calls the technique "a view-based exploit that only requires a user to view a malicious email" in a vulnerable client. It says the actors have been targeting and compromising Western government and commercial organizations through Zimbra since at least July 2025. The flaw, CVE-2025-66376 , is a stored cross-site scripting vulnerability in Zimbra's Classic UI. A crafted HTML email abuses CSS @import handling to execute JavaScript inside a...

Most of this week's trouble came dressed as something useful. A package stole data. A fake extension opened remote access. A safety app became spyware. An image gave hidden orders to an AI agent. Other threats hid in open systems, weak code, and normal network traffic. The threats change every week. Subscribe, and we’ll alert you when each new ThreatsDay Bulletin is out. The danger was easy to miss because it looked ordinary. Here is the full list:

Cybersecurity researchers have uncovered a sandbox escape vulnerability in Anthropic's Claude Cowork that makes it possible to break out of the confines of a Linux virtual machine (VM) within which the agent runs to read or write files anywhere on the Mac. Accomplish AI, which shared details of the vulnerability with The Hacker News ahead of publication, said about 500,000 macOS users running local Cowork sessions were affected prior to it being patched. It has been codenamed SharedRoot . "We connected a folder to a fresh Claude Cowork session, sent one short message, and watched the agent escape the sandbox," Oren Yomtov, principal security researcher at Accomplish AI, said . "From inside the VM, it reached the host Mac and read and wrote files all over it, far outside the folder we'd connected, with no permission prompt anywhere." With this level of access, the agent can access any data stored on the Mac via the user's account, including SSH ke...

The Chaos ransomware group ran its command-and-control through the victim's own browser. Cisco Talos on Thursday detailed msaRAT , the Rust implant behind it, found on a compromised Windows machine ahead of the encryptor. The implant never opens an outbound connection of its own. Its process talks to 127.0.0.1 and nothing else. It starts Chrome or Edge in headless mode and drives the browser over the Chrome DevTools Protocol, the browser's own debugging API. Every C2 message travels out from there through a WebRTC data channel relayed by Twilio's TURN service, so what a defender sees on the wire is a browser calling Cloudflare and Twilio. The attacker's own server address never appears at all. Chrome Does the Talking msaRAT looks for Chrome or Edge through environment variables first, then falls back to the registry for Chrome. If no matching browser is found, the CDP path is skipped. When it finds one, it starts the browser without a visible window using -...

An exposed Alibaba Cloud server has revealed a China-nexus operation that Group-IB tracks as JadeProx . The cluster has targeted government, healthcare, and education organizations across Asia and Latin America with a previously undocumented Windows loader called TriBack Loader. Group-IB found the server in mid-April 2026 in Alibaba Cloud's Singapore region; it was offline by the time the report published on July 23, 2026. Its bash history, phishing packages, post-exploitation tools, and webshell paths laid the operation out: active intrusions against a Vietnamese public hospital's medical imaging system and Malaysia's Ministry of Foreign Affairs, scanning and exploitation follow-up against Hong Kong education infrastructure, and a spear-phishing package addressed to the National Congress of Honduras. The operators reached the hospital's imaging server through webshells planted on an exposed Java management interface. One Loader, Four Builds TriBack Loader a...

Most people understand identity theft as an attacker stealing a real person's sensitive information and impersonating them. Synthetic identity fraud is much harder to catch. Instead of stealing a real identity, the attacker manufactures a new one, frankensteining together several real data points with fabricated ones to create a person who doesn't exist. Since no real victim monitors misuse, a fake identity can silently accumulate permissions and credibility over time before it's ever detected. This same principle has a largely unexplored parallel with Non-Human Identities (NHIs). Security teams are spending significant effort to protect NHIs from being stolen. Still, the machine-side equivalent of synthetic identity fraud is rarely discussed: identities that were never legitimately provisioned from the start. Following this approach, an attacker doesn't hijack an existing service account but instead fabricates one, blending real environmental attributes with fake o...

Cybersecurity researchers have shed light on a large-scale campaign that has turned compromised GitHub repositories into distributed attack infrastructure designed to target cPanel and WebHost Manager (WHM) instances. The activity involves malicious Packagist development versions spanning 10 packages associated with a legitimate PHP and DevOps developer, dinushchathurya, between July 12 and 13, 2026. The list of affected Packagist packages is below - dinushchathurya/nationality-list dinushchathurya/srilankan-divisional-secretariats dinushchathurya/srilankan-gn-divisions dinushchathurya/srilankan-local-authorities dinushchathurya/srilankan-mobile-number-validator dinushchathurya/srilankan-state-hospitals dinushchathurya/srilankan-universities dinushchathurya/uk-mobile-number-validator dinushchathurya/uk-post-code dinushchathurya/websmslk "The PHP libraries were not the execution path," Socket said in a statement. "Attackers had added...

Google on Thursday announced a new way for users to sign-in to their accounts by letting them take a selfie video. The selfie for sign-in , per the tech giant, is another option on top of existing recovery methods to log in to an account, including an email address or a phone number. The idea is to use a video selfie as a way to regain access if a user ever gets locked out or doesn't have access to their usual phone or computer. As part of the process, users are required to set up a selfie video by just looking into the device's camera and completing a "few short, guided head movements" to capture their face from different angles. Should users have any trouble signing in to their accounts with the selfie method at a later stage, they can just take another selfie to sign back in. "Selfie video compares the new video to the one you set up to confirm it is really you and help you get back into your account," Google said in a blog post shared with The Ha...

RefluXFS , a Linux kernel flaw disclosed on July 22 and tracked as CVE-2026-64600 , lets an unprivileged local user overwrite root-owned files on an XFS filesystem and gain persistent root access. Qualys said default installations of Red Hat Enterprise Linux and its derivatives, Fedora Server, and Amazon Linux can meet the conditions for exploitation. The company demonstrated the race against /etc/passwd and setuid-root binaries. The overwrite lands at the block layer. It survives a reboot and leaves the target's ownership, permissions, timestamps, and setuid bit untouched, so a modified setuid-root binary still runs as root. The fix was merged on July 16, and Linux vendors have begun shipping backported kernels. The patch traces the bug to Linux 4.11 in 2017: a Fixes: tag naming commit 3c68d44a2b49 and a stable backport request marked # v4.11 . Who is exposed Exploitation requires three conditions: The system runs Linux 4.11 or later without the RefluXFS fix....

Check Point has released security updates to address multiple vulnerabilities impacting Security Management and Multi-Domain Management (MDSM) products, including a critical flaw that has come under active exploitation in the wild . The security flaw, tracked as CVE-2026-16232 (CVSS score: 9.3), is an authentication bypass affecting the Check Point SmartConsole login process that allows an unauthenticated remote attacker to obtain an application login token and use it to authenticate with full administrative privileges. "Successful exploitation allows the attacker to modify security policies and security configurations," according to a description of the flaw in CVE.org. "Remote exploitation requires internet access to the Management Server IP address and a configuration that does not restrict Trusted Clients." Lotem Finkelstein, vice president of research at Check Point, said the company is aware of a handful of customers being targeted by this flaw, and t...

Beginning July 27, 2026, GitHub will cut public bug bounty payouts by at least half at every severity level. Critical findings will drop from $20,000-$30,000+ to a fixed $10,000, while its permanent invite-only VIP tier will pay $30,000 or more. Reports filed before that date, including those already in GitHub's growing triage queue, will retain the previous payout terms. GitHub said the changes are intended to reduce noise while giving established researchers faster responses, higher rewards, and closer access to its security engineering team. "You don't earn more by submitting more," the company said. "You earn more by submitting better." The public program is moving from flexible ranges to fixed payments: Low: $250, down from $617-$2,000 Medium: $2,000, down from $4,000-$10,000 High: $5,000, down from $10,000-$20,000 Critical: $10,000, down from $20,000-$30,000+ The Hacker News calculated that the new public rates are 50% lower f...

Cybersecurity researchers have disclosed details of a new local privilege escalation (LPE) vulnerability in snap-confine that an unprivileged user can trigger to obtain root access and gain complete control of a target environment. The high-severity flaw, tracked as CVE-2026-8933 (CVSS score: 7.8), impacts default installations of Ubuntu Desktop 24.04, 25.10, and 26.04. The disclosure comes as 442 security flaws in Linux have been publicized over the past three days. "The issue stems from a security hardening change that inadvertently introduced a race condition during sandbox initialization," Saeed Abbasi, head of Threat Research Unit (TRU) and director of product at Qualys, said. Snap-confine is a program used internally by snapd to construct the execution environment for snap applications. Snapd is the background service or daemon that manages snap packages on Linux systems. Snaps are nothing but a software packaging format devised by Canonical that allows an app...

Cybersecurity researchers have disclosed details of a now-patched vulnerability chain in the Adobe Acrobat Chrome extension that has over 314 million users, which, if exploited, could facilitate a silent hijack of a user's WhatsApp data. The shortcoming has been codenamed HermeticReader by Guardio Labs. It's officially tracked as CVE-2026-48294 (CVSS score: 7.4), with the vulnerability described as a case of universal cross-site scripting (UXSS)-class cross-origin data disclosure vulnerability. It affects all versions of the extension (ID: efaidnbmnnnibpcajpcglclefindmkaj ) prior to and including 26.5.2.2. Successful exploitation of the flaw can bypass the browser's same-origin policy and access data linked to the victim's session across origins. The only prerequisite is that it requires user interaction. A victim must be convinced into visiting a maliciously crafted URL or interact with a compromised web page that triggers the extension's vulnerable code ...

A high-severity security flaw impacting open-source developer platform Windmill has come under active exploitation in the wild, per VulnCheck. The vulnerability in question is CVE-2026-29059 (CVSS score: 7.5), a case of unauthenticated path traversal impacting Windmill's "get_log_file" endpoint ("/api/w/{workspace}/jobs_u/get_log_file/{filename}"). "The filename parameter is concatenated into a file path without sanitization, allowing an attacker to read arbitrary files on the server using ../ sequences," according to an advisory published by Windmill in March 2026. "The primary sensitive value exposed by this vulnerability is the SUPERADMIN_SECRET environment variable, readable via /proc/1/environ. When set, this secret can be used as a Bearer token to authenticate as a superadmin and execute arbitrary code through the job preview API." However, it's worth noting that SUPERADMIN_SECRET is not set by default, and for standalon...

Security leaders who build fast, visible paths to AI adoption are becoming the most valued partners in their organizations. AI governance done right gives security teams the visibility they need, employees the tools they want, and CISOs the strategic influence they have earned. According to McKinsey's State of AI report, 76 percent of employees now use AI in some capacity at work, up from 55 percent the year before. Writing assistants, coding copilots, meeting summarizers, AI-powered research tools: they are woven into daily work. Most were never reviewed by security. The conventional response is restriction. An application appears, the security team blocks it, employees find a workaround within days, and the cycle repeats. It comes down to speed: the approval path moves slower than the pace of AI releases. When the official route takes six weeks, and a workaround takes six minutes, most employees will choose the workaround. Technology gets adopted because people find it use...

OpenAI on Tuesday said a combination of its artificial intelligence (AI) models, including GPT-5.6 Sol and an "even more capable pre-release model," was behind the security incident that targeted Hugging Face's production infrastructure last week. The AI company said the models were operating with "reduced cyber refusals for evaluation purposes" that might otherwise limit their ability to conduct cyber attacks, adding it expects such incidents to "become more commonplace with the proliferation of increasingly cyber-capable models." Incidents like this may become more common. Subscribe, and we’ll alert you only when another major AI security story breaks. Describing it as an "unprecedented cyber incident" and one involving state-of-the-art cyber capabilities, OpenAI said it intends to conduct a thorough investigation in partnership with Hugging Face to get to the bottom of the matter. As part of an internal evaluation, the models...

The cycle is over. For years, cybersecurity followed a familiar pattern: defenses improved, attackers adapted, and the back-and-forth continued. Today, AI-equipped attackers are simply outpacing defenses. Most intrusions now bypass endpoint and malware-based detection entirely. The CrowdStrike Global Threat Report estimates around 79% of attacks are malware-free, as threat actors rely on credential theft and DLL side-load techniques to bypass host-level monitoring. Perimeter vulnerabilities compound this exposure; firewalls and VPN gateway breaches climbed 19% according to the latest Verizon Data Breach Investigations Report . Once an adversary gains access, breakout often occurs in seconds. Claude Mythos and similar models have further escalated operational pressure. These can rapidly discover and exploit previously unknown vulnerabilities, virtually closing the window from initial discovery to full compromise. Security practices must adapt to prioritize rapid containment and ...

German and US law enforcement have taken down the core infrastructure of Kratos , described by German investigators as one of the world's most widely used criminal phishing kits, and Indonesian authorities arrested the man they say developed and ran it. In a joint announcement on Monday, the Frankfurt public prosecutor's cybercrime unit (ZIT) and Germany's Federal Criminal Police Office (BKA) said they pulled more than 200 servers offline. Investigators estimate roughly 1,800 paying customers used Kratos to run about 15,000 phishing campaigns a month. Kratos harvested more than passwords. The kit was designed to steal the session cookie along with the login, and that cookie is enough to walk past two-factor authentication into the account as the user, the BKA said. ANY.RUN, which reverse-engineered the kit , found operators could pick one of two modes: a plain PHP page that only harvests credentials, or a Node.js reverse proxy designed to relay the login to Micros...

Cybersecurity researchers have discovered a NuGet typosquat that's unlike the typical information-stealing malware distributed via package registries: usual info-stealers: it's designed to rig live game results on Digitain. The package, named " Newtonsoftt.Json.Net ," masquerades as the Newtonsoft.Json library and is a trojanized fork. Seven versions of the package have been published to the NuGet repository: 11.0.4, 11.0.5, 11.0.7, 11.0.8, 11.0.9, 11.0.10, and 11.0.11. The package has been downloaded about 1,200 times to date. The package has been unlisted by its owner MagicalPuff96, meaning it will not be surfaced via a search on NuGet. However, the artifacts are still available for download from the registry. "The trojan rigs Digitain, an online betting platform, and in later generations, exfiltrates rigged round results to an attacker-controlled server, utilizing the header X-Seq-ApiKey: theperfectheist2025," according to JFrog . What's not...

A single invisible comment in an Azure DevOps pull request can turn a reviewer's own AI coding agent against them, driving it into projects the attacker has no rights to reach and quietly leaking what it finds. The flaw is in Microsoft's official Azure DevOps MCP server , and it works because one of its tools returns pull request descriptions without a prompt-injection guardrail the company had already applied to others. Agentic Runtime Security firm Manifold Security detailed the confused-deputy bug this week. Microsoft ships the server so AI agents can read and operate Azure DevOps for a user, across pull requests, pipelines, wikis, and work items, all with the user's own permissions. That is the whole problem: content other people wrote can become instructions the agent acts on. Azure DevOps PR descriptions accept Markdown, which allows HTML comments. In the web UI, an HTML comment (<!-- ... -->) renders as nothing, so a reviewer scrolling the description s...

Apple has moved to address a security flaw in its Hide My Email service that enabled users' real email addresses to be unmasked, effectively undermining the feature's privacy guarantees. 404 Media reported Tuesday that a fix for the issue was deployed by Apple on July 3, 2026, after more than a year, when it was disclosed to the company by Tyler Murphy, co-founder of EasyOptOuts. Hide My Email generates unique, random email addresses that forward messages to a user's personal email inbox automatically. By creating disposable email addresses, the idea is to safeguard user privacy and tackle unwanted spam. The feature requires a paid subscription to iCloud+ and was announced by Apple in June 2021. However, at the start of the month, details emerged of a flaw that made it possible to unmask a user's real email address hidden behind a Hide My Email address. The issue was first reported to Apple on June 13, 2025, with Cupertino unsuccessfully attempting to patch ...

Hidden text on a web page was enough to make Kiro , AWS's agentic coding IDE, rewrite its own configuration file and run an attacker's code on a developer's machine, with no approval step able to stop it. Intezer, in research with Kodem Security, found that a request as ordinary as asking Kiro to summarize a page could end in remote code execution. AWS has patched the issue and says it is covered by CVE-2026-10591. Kiro's safety model rests on a human clicking "allow." The agent can run shell commands, fetch URLs, and edit files, and the design assumes a developer reviews anything risky before it happens. That approval step is the security boundary, and the flaw let an attacker slip past it without the developer ever being offered a choice. The weak point was the file that tells Kiro which external tools to load. Kiro reads its list of Model Context Protocol servers, and the exact command used to start each one, from ~/.kiro/settings/mcp.json. When th...

Google's DeepMind on Tuesday announced the release of Gemini 3.5 Flash Cyber , a specialized artificial intelligence (AI) model built atop 3.5 Flash that's designed to discover, validate, and patch vulnerabilities quickly and efficiently. According to the tech giant, the model will be exclusively available to governments and trusted partners via CodeMender as part of a limited-access pilot program. CodeMender is an AI-powered agent for vulnerability discovery and patching that was unveiled by the company in October 2025. A Google DeepMind spokesperson told The Hacker News that there are plans to extend the model's capabilities to include red-teaming features and end-to-end enterprise defense. The lightweight model, per DeepMind, is both cost-efficient and highly capable alternative to large, costly cybersecurity-focused models. CodeMender can call upon 3.5 Flash Cyber "multiple times at high speed and low cost," allowing the AI agent to scan more code path...

A third SharePoint Server flaw patched by Microsoft as part of its Patch Tuesday update for July 2026 has come under active exploitation, per watchTowr . The vulnerability in question is CVE-2026-50522 (CVSS score: 9.8), a critical deserialization of untrusted data in Microsoft Office SharePoint that could allow an unauthorized attacker to execute code over a network. Microsoft credited DEVCORE researcher "splitline" with discovering and reporting the flaw. "In a network-based attack, an attacker authenticated as at least a Site Owner, could write arbitrary code to inject and execute code remotely on the SharePoint Server," Redmond said in an advisory released last week. "The attack vector is Network (AV:N) because this vulnerability is remotely exploitable and can be exploited from the internet. The attack complexity is Low (AC:L) because an attacker does not require significant prior knowledge of the system and can achieve repeatable success with the ...

Threat actors have been observed exploiting a now-patched high-severity Palo Alto Networks PAN-OS vulnerability as an entry point to deploy Qilin (aka Agenda) ransomware on victim environments. Arctic Wolf Labs said it investigated multiple intrusions in June 2026 that began with the exploitation of CVE-2026-0257 (CVSS score: 7.8), an authentication bypass flaw affecting the portal and gateway components of PAN-OS software. Successful exploitation of the flaw allows unauthenticated remote attackers to sidestep authentication and establish VPN sessions without valid credentials when authentication override cookies are enabled with specific certificate configurations. "Post-exploitation tradecraft varied across intrusions, from rapid encryption-only operations to full double-extortion, possibly suggesting multiple affiliates operating under the Qilin ransomware-as-a-service (RaaS) umbrella," the cybersecurity company said .

Zimbra has rolled out fixes to address multiple critical security issues, including a command injection flaw in the Simple Network Management Protocol (SNMP) monitoring component. As many as nine security vulnerabilities have been patched in Zimbra 10.1.20 . Topping the list is a command injection vulnerability in the SNMP monitoring component when SNMP notifications are enabled. Also patched are four cross-site scripting (XSS) flaws in the Classic Web Client - A stored cross-site scripting (XSS) vulnerability that could allow malicious attachment filenames to execute script under specific conditions. An XSS vulnerability where crafted fields could execute a malicious script under specific conditions. An XSS vulnerability where a crafted field could execute a malicious script when rendered. An XSS vulnerability where crafted attachments could execute a malicious script when rendered. Separately, fixes have been released for a mail forwarding restriction bypass (CV...

An Android app that can draw over other windows and write to shared storage can slip instructions to the AI agent driving that phone, in text no human eye will ever see. Two more steps, and the same app is running commands on the PC driving the agent. Researchers demonstrated that chain, plus six other attacks, against five open-source mobile agent frameworks: AppAgent, AppAgentX, Mobile-Agent-v3, Open-AutoGLM, and MobA. Everyone fell to at least six of the seven. The paper went up on arXiv on July 1 and was revised on July 14. The authors are at Simon Fraser University, the Chinese University of Hong Kong, Shandong University, and the Xingtu Lab at Chinese security firm QAX. Nothing here has a CVE, and first author Zidong Zhang told The Hacker News the team has no evidence of the techniques being used outside a controlled setting. The Hacker News checked all five frameworks and found the screenshot paths, the shell call, and the broadcast fallback that the paper describes st...

Every patch is a confession. The moment a vendor ships a security fix, the diff between the old code and the new code tells anyone watching exactly what was broken and where. Turn that diff back into a working exploit, and you can hit every system that hasn't updated yet. This is N-day exploitation, and it's always been a race: the vendor patches, the clock starts, and defenders try to deploy before an attacker finishes reverse-engineering the fix. For the last thirty-odd years, defenders usually won that race. Reverse-engineering a patch into a reliable exploit was slow, specialized work - usually requiring weeks of expert-level effort. Historically, the gap between a patch and a working public exploit spanned weeks, often months. The traditional playbook assumed you had at least a few weeks. You don't anymore. Not even close. Reverse-engineering a patch used to take weeks. Mythos does it in an hour. Anthropic's red team measured exactly that. Given not...

A cloud tenant using nothing but ordinary GPU access can push a data center's power draw up and down fast enough to threaten the grid it runs on, with no exploit and no break-in. That is the claim behind Bit2Watt , described by three Zhejiang University researchers in a paper accepted to CHES 2026 , the IACR's hardware-security conference, and the evidence splits in two: they measured the power modulation on real GPUs and simulated the grid destabilization it could cause. The technique inverts the usual grid-attack model: no compromised sensors, no malware on control systems, no stolen operator credentials, just a workload built to misbehave on purpose. It works because a GPU's power draw follows whatever it is computing. Saturate the tensor cores and current spikes; drop to idle, and it collapses. Toggle between those states on a schedule and you get a controllable power oscillation at the wall socket. The authors frame it as a blunt question: can "purely com...