Web Security | Breaking Cybersecurity News | The Hacker News

OpenAI has disclosed seven new incidents in which its artificial intelligence (AI) models exhibited misaligned behavior, including one case where its research model, after not having funds on an Ethereum test network, spent test cryptocurrency from a third-party's wallet without authorization to complete a coding task. The incident, the AI company said, involved an internal research model part of the GPT-5.6 Sol family. The incident occurred on June 26, 2026, and was discovered by OpenAI three months later. "An internal research model asked to update a cryptocurrency trading application used funds from a wallet that the task had not supplied to pay for a deployment on Sepolia, an Ethereum test network," OpenAI said . "It derived wallet keys from publicly known developer recovery phrases, found an account with test-network funds, and used that account to pay for the requested contract deployment." The assignment allowed the model to deploy compatible cont...

Candiru, an Israeli maker of spyware sold to governments, infected devices at the news network Al Jazeera through ordinary online ads, in attacks that needed no click or other action from the user, researchers at  the Citizen Lab  said Thursday. The hidden attack code ran the moment an ad loaded on the screen, and in some cases the ad loaded out of sight, in the background. The Citizen Lab, a research group at the University of Toronto, said this is the first public, technical confirmation that mercenary spyware, meaning hacking tools sold to governments, has reached real devices through targeted ads. The method had been described before from leaked company documents, but not shown in a live attack. The researchers recovered the spyware itself from one device, but not the exploit that breaks into the browser or app to install it. That first break-in step is described from how the attack was set up, not from captured code. How the Attack Worked The ads were placed by...

A long-running Traffic Distribution System (TDS) known as Help TDS to a Russian-language affiliate program called Hotejka , which appears to be a rebrand of a traffic monetization operation dubbed ROI777. Help TDS is assessed to be active since at least 2017, offering PHP code templates to partners and affiliates that are installed into compromised websites with an aim to direct unsuspecting users to tech support scams. These scams make use of full-screen browser manipulation and exit prevention techniques to serve bogus Microsoft Windows security alert screens instructing victims to contact a Windows Support number to address the problem. "Help TDS delivers tech support scams (TSSs) that direct users to call centers, where social engineering is used to convince them to install remote desktop tools," Infoblox said in a report shared with The Hacker News. "It's not simple credit card fraud: it's remote access, living-off-the-land malware operations. Once in,...

NGINX released versions 1.30.6 and 1.31.7 on October 14 to fix five vulnerabilities, one of them present in every release since May 2005. There are three memory bugs that F5, the company behind NGINX, says could crash a worker process or possibly let an attacker run code, and each requires a specific configuration to exploit. A fourth lets an attacker forge a TLS session that could get past client-certificate checks, though F5 calls the forgery non-deterministic and beyond the attacker's control. The fifth lets an HTTP/3 client slip extra bytes into a request that NGINX passes on to a backend. None of the five was in CISA's Known Exploited Vulnerabilities catalog as of October 15, and neither NGINX nor F5 reports attacks.  NGINX rates all five medium . F5 scores the three memory bugs 9.2, Critical, on CVSS v4.0, the TLS flaw 8.3, and the HTTP/3 flaw 6.9. Which Configurations Are Exposed Every one of the five needs a directive you added: an alias...

Next.js released an  out-of-band security update  on October 14 that fixes two critical flaws allowing unauthenticated remote code execution, along with a high-severity denial-of-service flaw. One code-execution flaw is in React Server Components, and the other in librsvg, the open-source library Next.js uses through the sharp package to handle SVG images. The fixes are in Next.js 16.4.1 and 15.5.28, the current Active and Maintenance long-term support (LTS) releases. What Was Fixed The update covers three separate issues. Vercel, which develops Next.js, rated both code-execution flaws critical, each with a CVSS score of 9.5. The table lists the versions each issue affects and the version that fixes it. Issue Affected versions Fixed in Who is exposed Code execution through use cache ( GHSA-673q-mjm6-r5xr ...

Drupal core has a critical flaw in its Field UI module that could let an attacker run code on the site's server, and fixes were shipped on October 14 in Drupal 11.4.10, 11.3.19, and 10.6.19. An attacker cannot reach the flaw through core alone, the Drupal Security Team said in  its advisory . Several exploit paths exist through contributed modules, and at least one known path needs no login. Tracked as CVE-2026-108097 , the flaw sits in Field UI, the core module behind the admin pages where site builders manage fields and how they display. Drupal rates it 19 out of 25 on its own risk scale, the level the project calls critical. The advisory does not name the contributed modules that open a path to the flaw, and custom code may also expose it. Drew Webber of the Security Team and Octavio Galland are credited with reporting it. What to Do Now The fixed release for each supported branch: Drupal 11.4.x: update to  11.4.10 Drupal 11.3.x: update to  11.3.19 ...

Threat actors are exploiting a critical security flaw in Langflow, according to an alert from Previdian. Langflow is an open-source platform that allows developers to build artificial intelligence (AI) applications and agentic workflows. In February 2025, DataStax, the parent company of Langflow, was acquired by IBM. The vulnerability in question is CVE-2026-105740 (CVSS score: 9.9), a case of authenticated remote code execution impacting versions prior to 1.9.0. "Any authenticated Langflow user can achieve Remote Code Execution (RCE) on the server by adding an MCP server with the 'stdio' transport," reads an advisory released last month. "The user-supplied command field is passed directly to bash -c "exec {command}" with zero validation, no allowlisting, and no sandboxing. The command executes immediately when the server list is fetched. Additionally, the env field allows arbitrary environment variable injection (e.g., LD_PRELOAD, PATH ove...

Singapore-based Ahsay Backup has released security updates to address two security flaws in AhsayCBS that have been exploited by bad actors to drop web shells and XMRig cryptocurrency miners. The vulnerabilities in question are CVE-2026-105133 and CVE-2026-105134 , which could be chained to sidestep authentication and execute arbitrary commands on affected systems. In a report published earlier this month, Huntress said it observed threat actors combining the two flaws to achieve unauthenticated remote code execution and deploy web shells and cryptocurrency miners on exposed systems. Although the vulnerabilities were initially assumed to be addressed in AhsayCBS versions 9.15.2.59 and 10.3.4.0 released in August 2026, the software company has now confirmed that the patches did not completely plug them. The fixes are available in the following versions - AhsayCBS v10.3.4.45 hotfix AhsayCBS v9.15.2.70 hotfix "Given the severity of these risks, we strongly urge ...

Attackers exploited a flaw in an optional Nextcloud app,  Collabora Online - Built-in CODE Server , to plant web shells on Nextcloud servers. Three administrators  reported the issue on GitHub , with the first logged attack on September 10. A web shell is a small file that executes any command an attacker sends via a web request, with the rights of the web server account. The flaw allows an unauthenticated request to run a command on the server, but only when the app's local document server is not running, such as during a restart or upgrade, or after a crash. Collabora fixed it in  version 26.4.303  on September 11 and on October 14 published  an advisory  with the identifier  CVE-2026-110447 , warning that turning off the app does not help: it must be updated or removed. Who Is Affected The app, published by Collabora under the package name richdocumentscode, downloads a copy of Collabora Online Development Edition (CODE) and runs it ...

A flaw in how curl handles HTTP/2 uploads lets a malicious server receiving an upload corrupt the client's memory, and the researcher who reported it turned that into code execution on the uploading machine. The curl project fixed it on October 14 in  curl 8.23.0 . Taking over a client needs exact knowledge of its build and memory layout, and curl expects a real attack to end in a crash instead. The flaw, a use-after-free tracked as  CVE-2026-92392 , affects curl and libcurl 8.12.0 through 8.22.0, both the command-line tool and the library that many other programs use for transfers. curl rates it High on its own four-level scale, its first High-severity flaw since the  SOCKS5 heap overflow of October 2023 . The project cut its release cycle by two weeks to ship the fix. The attack runs from the server side. The client has to upload a file of roughly 135KB over HTTP/2 to a server controlled by the attacker, and the server's certificate must pass the client's checks....

A team at security firm XBOW chained three then-unknown Google Chrome bugs into a single attack that runs code on a Windows computer when the user only loads a web page, the company said. Its researchers built the attack in six days by directing AI agents running OpenAI's publicly available GPT-5.6 Sol model. Google has since confirmed all three bugs,  fixed them , and paid XBOW its top $250,000 reward for a full Chrome exploit. The attack is a full chain: a set of bugs used together to break every layer of Chrome's protection, from a web page all the way to the computer. Chrome runs web page code inside two nested sandboxes, the barriers that stop a hacked page from reaching the rest of the machine. The chain crossed both. The result was code running outside Chrome on Windows, with no need for the user to click or download anything. All three bugs are now fixed, and XBOW says they were never used against Chrome users. The two that carry public identifiers were fixed ...

The bug was simple: the login let you choose where your sign-in code went. Proving it on a live website was the hard part. TL;DR During an authorized assessment, an AI pentesting agent found a critical authentication bypass on a major insurer's production website, on its own. One login request let the user choose where the one-time passcode went. The attack was the easy part. Proving it on a live site took web context built over a decade. The Login Worked. Then the Agent Stopped. Reflectiz's AI pentesting agent was mapping the login flow of a major insurer's production website, which has millions of users. The agent requested a login code and directed it to an email inbox it controlled. The code arrived. The login went through. Its guardrails halted the test after the login succeeded.  The agent had discovered a critical-severity authentication bypass on its own. Reflectiz reported it to the insurer. The Flaw: The Request Chose Where the Code Went ...

Two of 31 newly disclosed flaws in Progress Software's  MarkLogic Server  database could let an account given only some management privileges take administrator control of the server, the company says. Progress published all 31 on October 12 and scored five of them critical. Fixes have been available since September 10, in  MarkLogic 11.3.7  and  MarkLogic 12.1.0 . Affected releases, and the release that fixes each: MarkLogic 11.x before 11.3.7: upgrade to 11.3.7 MarkLogic 12.x before 12.1.0: upgrade to 12.1.0 MarkLogic 11.3 is the long-term support (LTS) line. The 12 line, which Progress calls an innovation release, has no patched 12.0 release. Its fix is 12.1.0, which also adds new features. Installing the fixes for the 10 MarkLogic flaws Progress disclosed in August is not enough. Those fixes arrived in 11.3.6 and 12.0.3, and both releases are affected by the new flaws. Progress's CVE records and its upgrade notes for the two releases do n...

GitLab has released patches for a maximum-severity security flaw impacting its Community Edition (CE) and Enterprise Edition (EE) software that could result in arbitrary code execution. "GitLab has remediated an issue that, under certain conditions, could have allowed an unauthenticated user to execute arbitrary code on the GitLab server via a specially crafted regular expression in a CI/CD configuration, due to insufficient regular expression validation," the company said in a Monday advisory. The flaw, tracked as CVE-2026-97229 , has been described as an integer overflow issue. It carries a CVSS score of 10.0. It affects the following versions - All versions from 19.2 before 19.2.8 All versions from 19.3 before 19.3.5, and All versions from 19.4 before 19.4.2 Patches for the flaw are available in versions 19.4.2, 19.3.5, and 19.2.8. Also addressed are three high-severity bugs - CVE-2026-92883 (CVSS score: 8.7) - A cross-site scripting vulnerability...

Threat actors have been observed exploiting a recently disclosed critical security flaw in SonicWall Secure Mobile Access (SMA) 1000 appliances within days of public disclosure. The vulnerability, tracked as CVE-2026-102255 (CVSS score: 10.0), is a case of pre-authenticated server-side request forgery (SSRF) that could allow a remote attacker "to direct the appliance to issue requests on their behalf and reach internal functionality and perform unauthorized operations." Patches for the flaw were released by SonicWall on October 6, 2026, with versions 12.4.3-03670 (platform-hotfix) and 12.5.0-03082 (platform-hotfix). The vulnerability has since started witnessing active exploitation attempts, with first signs detected by threat intelligence firms Defused Cyber and Previdian on October 8, 2026. Data from Previdian shows that a total of 135 attempts have been observed to date from two unique IP addresses located in the U.S. SonicWall has yet to update the advisory ...

A maximum-severity security flaw impacting the Bricksforge plugin for WordPress has come under active exploitation in the wild, according to Patchstack. The critical vulnerability in question is CVE-2026-85097 (CVSS score: 10.0), which allows unauthenticated attackers to upload and execute arbitrary PHP code on vulnerable WordPress websites, resulting in remote code execution (RCE). The security flaw impacts Bricksforge versions up to and including 3.1.8.9. Patches have been released in version 3.1.8.10. Bricksforge is a WordPress plugin that's advertised as a "Swiss Army knife" for Bricks, a visual website-building tool for WordPress, with support for animations, advanced forms, and scrolling video sequences. WordPress security company Patchstack said it detected exploitation efforts targeting the flaw on October 7, 2026, at 9:47 p.m. UTC. "Bricksforge validates a file's MIME type when it is first uploaded. When a form is later submitted, however, it ...

Cybersecurity researchers have disclosed details of a two-pronged artificial intelligence (AI)-augmented initial access campaign targeting WordPress sites, e-commerce platforms, and critical infrastructure entities in Ukraine. The activity, which brings together commodity offensive tools like Cobalt Strike with an AI orchestration layer powered by Anthropic's Claude Code, has been attributed to a Russian-speaking, financially motivated initial access broker codenamed CyberXero . The threat actor is assessed to be active since at least July 2026. "The operation runs two pipelines at once: a global, automated campaign against WordPress and e-commerce platforms, and a curated, manual campaign against Ukrainian energy and critical infrastructure," SOCRadar said in a report published last week. The global campaign specifically focuses on known security flaws like wp2shell (CVE-2026-63030 and CVE-2026-60137) and CVE-2026-4815 , an SQL injection vulnerability impactin...

Anthropic on Friday said it's cutting off live internet access for all its internal evaluations until further notice following the discovery of new incidents in which its artificial intelligence (AI) models exhibited misaligned behavior and targeted real websites. The AI company said it identified four broad categories of unintended model actions during evaluations and internal use of Claude - Claude Mythos Preview exploiting SQL or command injection flaws in unspecified third-party software to run commands on a university server, either because its own tools were intentionally limited or because an outside service it needed was unavailable, causing it to use other tools hosted on a third party's site to complete the task. Claude Haiku 4.5 and a non-frontier research model submitting a sensitive form on a real website when it was not authorized to do so. This occurred in scenarios where instructions were ambiguous or due to environment misconfigurations that preve...

Threat actors have been observed exploiting two recently disclosed flaws in the AhsayCBS backup utility to seize control of affected devices and deploy web shells and XMRig cryptocurrency miners. Details of the flaws are below - CVE-2026-105133 (CVSS v4 score: 5.5) - An improper authentication vulnerability in the checkSysPwd() function in the "com/ahsay/obs/api/ApiStructsAction.java" component. CVE-2026-105134 (CVSS v4 score: 9.3) - An operating system command injection vulnerability in the Replication Receiver component. A remote attacker could chain the two vulnerabilities to bypass authentication and execute arbitrary commands on affected systems. It's worth noting that CVE identifiers for these flaws were not published until October 4, 2026. According to Huntress, exploitation efforts aimed at the two flaws began on October 7, 2026, at 11:20 p.m. UTC, with unidentified threat actors weaponizing them to achieve remote code execution on impacted hosts...

The U.S. Cybersecurity and Infrastructure Security Agency (CISA) on Thursday added five security flaws to its Known Exploited Vulnerabilities ( KEV ) catalog, following their abuse by a China-linked threat actor known as Flax Typhoon. The vulnerabilities in question are listed below - CVE-2015-3306 (CVSS score: 10.0) - An improper access control vulnerability in ProFTPD that could allow remote attackers to read and write to arbitrary files via the site cpfr and site cpto commands. CVE-2021-3199 (CVSS score: 9.8) - A path traversal vulnerability in ONLYOFFICE Docs that can occur when JSON Web Token (JWT) is used, via a "/.." sequence in an image upload parameter and could allow for remote code execution. CVE-2023-22894 (CVSS score: 7.2) - A cleartext storage of sensitive information vulnerability in Strapi that could allow an attacker with access to the admin panel to discover sensitive user details via the query filter. CVE-2016-3081 (CVSS score: 8.1) - A...