Open Source | Breaking Cybersecurity News | The Hacker News

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...

In an ironic twist, open-source artificial intelligence (AI) platform Hugging Face revealed that it was the victim of a hack perpetrated by an autonomous AI agent system. The company said it detected and responded to the incident targeting its production infrastructure earlier last week. "We identified unauthorized access to a limited set of internal datasets and to several credentials used by our services," the company said in a statement. While an investigation into the intrusion remains ongoing, Hugging Face said it has found no evidence that the AI agent tampered with public, user-facing models, datasets, or Spaces, and its own software supply chain. The starting point of the attack was the data processing pipeline itself, with a malicious dataset abusing two code execution paths, viz., in its remote code dataset loader and a template injection in a dataset configuration, to run code on a processing worker.

Cybersecurity researchers have flagged a new software supply chain attack codenamed SleeperGem targeting the Ruby ecosystem after three malicious gems were published to RubyGems with the end goal of serving additional payloads. The rogue gems are listed below - git_credential_manager (versions 2.8.0, 2.8.1, 2.8.2, 2.8.3) - Published on July 18, 2026 Dendreo (versions 1.1.3, 1.1.4) - Published on October 14, 2017 fastlane-plugin-run_tests_firebase_testlab (version 0.3.2) - Published on February 06, 2018 "Each malicious release is a loader," StepSecurity said in an analysis. "It fetches a second stage from an attacker controlled Forgejo host, checks whether it is running in a build system and skips if it is, and on a developer machine it drops a native daemon and installs persistence." One aspect of the attack that immediately stands out is that "git_credential_manager" impersonates the official Microsoft Git Credential Manager, while t...

Eleven bytes will make an unpatched OpenSSL server set aside up to 131 KB of memory for a message that never arrives. On the glibc systems Okta tested, that memory is gone until the process restarts. OpenSSL shipped the HollowByte fix in June with no CVE, no advisory, and no changelog entry pointing at it. Okta's Red Team, which reported the denial-of-service bug and named it, published the details on Thursday. The fixed releases are OpenSSL 4.0.1, 3.6.3, 3.5.7, 3.4.6, and 3.0.21 , all dated June 9. Every release on those branches before the fixed ones has it. Nothing in a normal patch pipeline will point you at them: there is no identifier for a scanner to match and no advisory to read. The flaw is that OpenSSL took the attacker's word for it. Every TLS handshake message carries a 4-byte header, three bytes of which declare how long the body will be. Older versions grew the receive buffer to that declared size the moment the header landed, before a single byte of the bo...

Give an AI assistant a memory and access to your inbox, and you hand an attacker a way to rewrite what it thinks it knows about you. A single email can trick that agent into saving a false "fact" about the user, hide the change, and quietly steer its answers in later sessions. When it works, the person reads an ordinary-looking reply and never learns their assistant was tampered with. The researchers named the attack  stealth memory injection  and built a tool that writes the emails automatically. The paper, "When Claws Remember but Do Not Tell,"  landed on arXiv on 6 July 2026 . First, what these assistants do A personal agent is an AI assistant that sticks around. Instead of forgetting everything when a chat ends, it keeps notes about you in files: your preferences, your contacts, and what you asked it to do. It reads those notes at the start of every new session, which is why it feels like it knows you. Many of these agents can also act for you, readin...

Details have emerged about three now-patched security flaws in the OpenClaw personal artificial intelligence (AI) assistant that, if successfully exploited, could enable credential theft, privilege escalation, and arbitrary code execution on the host. A brief description of the high-severity vulnerabilities is as follows - GHSA-hjr6-g723-hmfm (CVSS score: 8.8) - An operating system command injection and an incomplete list of disallowed inputs vulnerability impacting the host execution environment filtering mechanism that could allow for executing or persist actions beyond the caller's intended authorization. GHSA-9969-8g9h-rxwm (CVSS score: 8.8) - An operating system command injection and an incomplete list of disallowed inputs vulnerability impacting the host execution environment filtering mechanism that could allow for executing or persist actions beyond the caller's intended authorization. GHSA-575v-8hfq-m3mc (CVSS score: 8.4) - A path traversal and link f...

A single wrong variable on one line in XQUIC, Alibaba's QUIC and HTTP/3 library, lets any remote client crash the server with a short burst of completely legal traffic. There is no patch. FoxIO researcher Sébastien Féry  disclosed the flaw on July 8  and nicknamed it XRING. He says it needs no login and no malformed packets: about 260 bytes of ordinary QPACK traffic takes the server process down. XQUIC is open-source, so the risk is not Alibaba's alone: any server that embeds it and serves HTTP/3 with the default QPACK settings is exposed. That includes Tengine, Alibaba's Nginx-based web server, which FoxIO says fronts the company's cloud and CDN on sites including Taobao and Alipay. Every release through v1.9.4, the latest, is affected. There is no fixed release and no CVE as of July 10. Until a fix ships, operators can set SETTINGS_QPACK_MAX_TABLE_CAPACITY to 0, which turns off QPACK's dynamic table, or drop HTTP/3 support entirely. The bug lives in how H...

GitHub has officially announced the release of npm version 12 with install scripts disabled by default, along with deprecating granular access tokens (GATs) designed to bypass two-factor authentication (2FA). The Microsoft-owned subsidiary noted that the following npm install behaviors that used to run automatically before have been made opt-in - allowScripts defaults to off, meaning dependency lifecycle scripts (i.e., preinstall, install, postinstall) and implicit node-gyp builds no longer run unless explicitly allowed. --allow-git defaults to none, meaning --allow-git defaults to none: Git dependencies (direct or transitive) are no longer resolved unless explicitly allowed. --allow-remote defaults to none, meaning dependencies from remote URLs (e.g., https tarballs) are no longer resolved unless explicitly allowed. To review and approve trusted scripts, users are now required to run: "npm approve-scripts --allow-scripts-pending," then commit the resulting a...

Everyone seems to have announced a clearinghouse over the past few weeks. We did too. Ours is called Athena , and the main thing that sets it apart is that it was already real and running when we announced it — built quietly months earlier, heads down, taking findings and shipping fixes, because customers kept asking us to. We only announced it now because everyone else started announcing theirs, and staying quiet started to look like something it wasn't. The others arrived louder and, as far as anyone outside the press releases could tell, didn't exist yet. Here's the part none of those announcements will tell you: the clearinghouse is the least important thing to build. When a project we'd deliberately kept private, a  five-billion-dollar press release , and  the White House all reach for the same word inside a few weeks, that's not a trend. Trends are optional. This is the shape of a problem changing under everyone at once. So let me explain why these thin...

Ask an AI coding agent to scan open-source code for security holes, and it might run the attacker's code on your own machine instead. That is the finding in a  proof-of-concept published Wednesday by the AI Now Institute, an attack it calls " Friendly Fire. " It works against Anthropic's Claude Code and OpenAI's Codex when either is running in an autonomous mode that approves its own commands. It hijacks the exact job these tools are sold for: checking untrusted third-party code for problems. Instead of catching the threat, the agent becomes the way in. Researchers Boyan Milanov and Heidy Khlaaf tested two setups, each a stock install with the autonomous mode switched on: Claude Code (CLI 2.1.116, 2.1.196, 2.1.198, 2.1.199) on Claude Sonnet 4.6, Sonnet 5, or Opus 4.8 OpenAI Codex (CLI 0.142.4) on GPT-5.5 Claude Code's "auto-mode" and Codex's "auto-review" use a classifier to run commands the agent judges safe, pausing ...

The safety check that is supposed to stop an AI coding agent from running a dangerous command can be walked straight past using a shell trick that has been public for decades. New research from  Adversa AI , which is named the bypass GuardFall , found it works against ten of the eleven popular open-source coding and computer-use agents the firm tested. Only one, "Continue," was built to defend against it. Why does it matter? These agents run shell commands with your full account access. Point one at a booby-trapped repository or software package, and a hidden instruction can quietly run a command that wipes files or steals the secrets your account can reach, from SSH keys and cloud credentials to anything sitting in your home folder. How does it get past the guard? Most of these agents try to stay safe by checking each command against a blocklist of dangerous patterns before running it. The flaw is that they check the command as plain text, while bash rewrites that t...

A public proof-of-concept is now out for CVE-2026-55200 , a critical flaw in libssh2 that lets a malicious or compromised SSH server trigger memory corruption on a connecting client, with possible code execution. No credentials, no user interaction. The bug affects every release up to and including 1.11.1 and carries a CVSS 4.0 score of 9.2 . libssh2 is a client-side SSH library, not a server. That distinction matters. It is embedded in curl, Git, PHP, backup agents, firmware updaters, and a long tail of appliances. Anything that links it and reaches out to an untrusted SSH endpoint is a potential target. Many of those copies are statically linked, so a distro package update will not touch them, and you may not know they are there. How the bug works The flaw lives in ssh2_transport_read() in transport.c, the function that parses incoming SSH packets during the handshake. It read the attacker-controlled packet_length field and rejected only values belo...

Cybersecurity researchers have flagged a new class of CI/CD workflow weakness that allows attackers to hijack workflows and compromise open-source supply chains. The "critical exploitable pattern" has been codenamed Cordyceps by Novee Security. The issue can allow full attacker control of repositories at dozens of the largest organizations worldwide, including Microsoft, Google, Apache, and Cloudflare. "The flaw is exploitable by any unauthenticated user," Elad Meged, founding engineer and security researcher at Novee Security, said . "No org membership or special privileges; a free account is enough to forge approvals, push code, or steal credentials." The penetration-testing company's scan of about 30,000 high-impact repositories has revealed more than 300 to be fully exploitable, enabling attacker-controlled code execution, credential theft, and supply chain compromise, which can have severe downstream impacts. The core of the problem tri...

Security firm AIR built a fake AI agent skill, pushed it through a popular skill marketplace and an Instagram ad, and says it reached roughly 26,000 agents, including some on corporate accounts. Every skill security scanner the firm tested it against marked it safe. The payload was harmless by design: it collected the user's email address and did nothing else. The point was to show that none of the signals people lean on to trust a skill caught it: not the scanners, not the GitHub stars, not the open-source reputation. A skill is a bundle of instructions an agent loads into its own context and follows with roughly the authority of a user prompt. That trust is the whole problem, and it is the reason skill-scanning tools exist in the first place. The skill, named  brand-landingpage , claimed to build a landing page using Google's Stitch design tool, aimed squarely at non-technical users. To make it look credible, AIR went after two trust signals: GitHub stars an...

OpenAI on Monday said it's releasing an improved version of its GPT‑5.5‑Cyber model to trusted defenders as part of the Daybreak initiative  the artificial intelligence (AI) company announced last month. Calling GPT‑5.5‑Cyber its "strongest model yet for finding and helping patch software vulnerabilities," OpenAI said the model can "sustain deeper analysis across large codebases" to identify security issues, validate them in a controlled environment, and develop and test patches. In tandem, the tech upstart is releasing an update to the Codex Security plugin⁠ to speed up the process of discovering and patching vulnerabilities in existing systems, alongside preventing new vulnerabilities from entering production codebases. "Developers can run deep scans or review recent changes, generate reports with severity, affected code locations, validation evidence, and remediation guidance, trace attack paths, build threat models, validate findings, and genera...

Google has set September 30, 2026, as the day it begins enforcing  Android developer verification  in the first four countries, and the major device-maker app stores are in from the start. On that date, certified Android phones in Brazil, Indonesia, Singapore, and Thailand will block normal installs of apps whose developers have not registered an identity with Google, whether the app comes from Google Play or the stores run by Samsung, Xiaomi, OPPO, vivo, Honor, and Transsion. Certified devices are the ones that ship with Google's services and Play Protect, which, by F-Droid's count, is more than 95 percent of Android devices outside China. Most users will not notice, which is the point. Apps from verified developers keep installing as before. The friction lands on apps from developers Google has not verified, and is hardest on the independent and open-source channels, built on not needing Google's permission to ship. Developers distributing through those stores ne...

As many as 145 npm packages associated with the Mastra namespace ("@mastra/*"), a popular open-source JavaScript and TypeScript framework for building artificial intelligence (AI) applications, have been compromised as part of a software supply chain attack codenamed easy-day-js , per findings from Endor Labs , JFrog , OX Security , SafeDep , Socket , StepSecurity , and Synk . "A single npm account (ehindero) mass-published more than 140 malicious packages across the Mastra scope within a short window on 2026-06-17," Socket said. The infected packages themselves do not include malicious code. Instead, it's introduced by means of a third-party library named "easy-day-js" that has been added to each package's dependency list in what has been described as an automated publishing campaign spanning 88 minutes. In its analysis, SafeDep described "easy-day-js" as a clone of the "dayjs" date library that downloads and runs a ...

A default low-privilege account on a LiteLLM proxy can climb to full admin and run code on the server by chaining three vulnerabilities, researchers at Obsidian Security disclosed LiteLLM is a widely deployed open-source AI gateway that brokers calls to more than 100 model providers behind one OpenAI-compatible interface. A server takeover exposes every provider key it holds, the secrets that decrypt its stored credentials, and every prompt and response passing through it. Obsidian rates the full chain CVSS 9.9, in the Critical range. BerriAI , the maintainer, included the complete fix set in LiteLLM v1.83.14-stable, which GitHub lists as released May 2. Upgrade to that release or later to close the three-CVE chain. The three bugs The first link is CVE-2026-47101 , an authorization bypass. When a regular user (an internal_user) generates a virtual API key, LiteLLM stores the caller-supplied allowed_routes field without checking it against the user's role. The field is...

A high-severity security flaw in Langflow, an open-source low-code platform to build artificial intelligence (AI) applications, has come under active exploitation in the wild, according to findings from VulnCheck. The vulnerability in question is CVE-2026-5027 (CVSS score: 8.8), a case of path traversal that could allow an attacker to write files to arbitrary locations. "The 'POST /api/v2/files' endpoint does not sanitize the 'filename' parameter from the multipart form data, allowing an attacker to write files to arbitrary locations on the filesystem using path traversal sequences ('../')," Tenable, which discovered the flaw, said in an alert released in late March 2026. The cybersecurity company said it attempted to contact the project maintainers three times in January and February 2026, before disclosing details of the issue on March 27. Caitlin Condon, vice president of security research at VulnCheck, said in a LinkedIn post that the v...

On June 9, Anthropic released Claude Fable 5 , the most capable model it has ever made, generally available. It also did something unusual: it shipped one model as two products, split not by capability but by a layer of safety classifiers. Fable 5 goes to the public. Its twin, Claude Mythos 5, the same underlying model with the cyber safeguards lifted, stays locked to a vetted group of cyber defenders and critical infrastructure operators. Anthropic calls Mythos 5 the strongest cybersecurity model in the world. The practical difference is this: Fable 5 routes flagged cyber, biology, chemistry, and distillation requests to the weaker Claude Opus 4.8, while Mythos 5 keeps the cyber capabilities available for vetted users. Both models cost $10 per million input tokens and $50 per million output tokens, less than half the price of the earlier Mythos Preview, and Fable 5 is available through the Claude API now. It is included on Pro, Max, Team, and seat-based Enterprise plans at no...