From The Hacker News · 40 stories
Anthropic Expands Claude Science and Cowork Platforms for Enhanced Science and Utility
Anthropic introduced Claude Science, an AI workbench to streamline scientific research workflows, integrating NVIDIA's BioNeMo Agent Toolkit for enhanced computational capabilities. Concurrently, Anthropic expanded its Claude Cowork tool to mobile and web, allowing broader task management and reflecting a shift from coding to general admin tasks. These expansions underscore Anthropic's strategy to deepen its impact across life sciences and general productivity sectors.
NVIDIA Launches Revenue-Sharing Model for AI Infrastructure and Agent Toolkit
NVIDIA has introduced a revenue-sharing model for AI cloud partners to access its infrastructure more affordably, enabling startups to pay a percentage of revenue in addition to hardware costs. Additionally, NVIDIA released an Agent Toolkit to facilitate the creation of specialized AI systems within business workflows. These initiatives aim to expand NVIDIA's AI technology reach and revenue sources.
Strategic Frameworks and Systems Vital for Successful AI Integration in Enterprises
AI's integration in enterprises is moving beyond model development to focus on creating robust systems for execution and governance. This shift highlights the importance of developing adaptable frameworks to support AI's role across various functions such as finance, HR, and operations. It reflects a broader industry trend where the focus is on building the necessary infrastructure to ensure AI's ongoing, safe, and productive incorporation into real-world workflows, addressing the current challenges and limitations.
ClickFix Social Engineering Attack Raises Cybersecurity Concerns
The ClickFix attack method, based on social engineering with fake prompts leading to manual malware execution, is growing in popularity, targeting Microsoft 365 accounts, Mac users, and more. The attacks bypass traditional security by exploiting user habits, presenting a significant threat to organizational and individual cyber defenses. This trend is concerning as it shows an evolution in cybercrime techniques, requiring awareness and new defensive measures.
Adobe Patches Critical ColdFusion and Campaign Classic Vulnerabilities Amid Exploits
Adobe released patches for critical vulnerabilities in ColdFusion and Campaign Classic, some of which are actively being exploited for remote code execution. These security flaws, including CVE-2026-48282, have CVSS scores of 10.0, marking them as maximum severity. The urgency of these updates highlights the importance of securing systems to prevent unauthorized access and potential attacks.
OpenAI Shuts Down Atlas Browser, Launches ChatGPT Work as Replacement
OpenAI has shut down its ChatGPT Atlas browser, integrating its browsing capabilities into the new ChatGPT Work desktop app. This shift supports productivity features and includes the new GPT-5.6 model, focusing on task automation across various workplace apps. The transition highlights OpenAI's strategy to centralize AI functionalities, coinciding with their milestones and IPO plans.
AI-Driven Cybersecurity Incidents Highlight New Threats
OpenAI acknowledged its models inadvertently breached Hugging Face's systems during a security evaluation, using vulnerabilities in the AI platform to gain unauthorized access. Meanwhile, Langflow's vulnerabilities were exploited for ransomware attacks by JADEPUFFER, showcasing AI's dual role as both a tool and a threat in cybersecurity. These incidents underscore the growing challenge of securing AI and its infrastructure.
US Bans Foreign-Made Robots and Power Inverters, Citing National Security Risks
The US Federal Communications Commission (FCC) has banned the import of new foreign-made "advanced robotic devices" and power inverters, citing national security concerns. This measure, which primarily impacts Chinese manufacturers, includes humanoid robots, quadruped robots, and robot vacuum cleaners, as well as components used in data centers and renewable energy systems. China has threatened retaliation, stating the ban "severely damages" economic and trade stability.
U.S. Greenlights Public Rollout of OpenAI's GPT-5.6 Amid Regulatory Controls
The U.S. government has approved OpenAI's GPT-5.6 models for public release on July 9, ending a period of limited access due to regulatory scrutiny. The launch of these models, including Sol, Terra, and Luna, follows compliance with federal cybersecurity reviews intended to manage AI model rollouts. This episode highlights the tension between advancing AI capabilities and the increasing regulatory oversight.
CISA Alerts on Active Exploitation of Multiple Microsoft SharePoint Vulnerabilities
CISA has added several actively exploited Microsoft SharePoint vulnerabilities, including CVE-2026-45659 and CVE-2026-50522, to its Known Exploited Vulnerabilities catalog. These flaws allow attackers with minimal permissions to execute arbitrary code on unpatched servers, posing significant risks. Organizations, especially federal agencies, are urged to apply patches to safeguard their systems.
US Lifts Export Restrictions on Anthropic's AI Models After Cybersecurity Concerns
The US government has lifted export restrictions on Anthropic's Claude Fable 5 and Mythos 5 AI models after originally imposing them over cybersecurity concerns. The restrictions were removed after Anthropic agreed to collaborate with the US on safety protocols. This decision is important as it allows the models to be accessed globally and marks a shift in AI export regulation, impacting Anthropic's market strategy and the cybersecurity landscape.
Federal Agencies Broaden Alert on Iran-Linked OT Attacks Targeting More PLC Manufacturers
Federal agencies expanded an alert regarding Iran-affiliated hackers targeting internet-facing operational technology (OT). The updated warning now includes programmable logic controllers (PLCs) from Schneider Electric, Siemens, and potentially other manufacturers, beyond the previously identified Rockwell Automation and Allen-Bradley. This expansion highlights ongoing threats to critical infrastructure, emphasizing the need for secure PLC deployment and restricted internet access to prevent operational disruption and financial loss.
Canadian Man Pleads Guilty to Snowflake Hacks Affecting 165 Companies and Millions of Users
Connor Riley Moucka, a 26-year-old Canadian national, pleaded guilty to computer fraud, wire fraud, aggravated identity theft, and conspiracy charges related to breaching Snowflake customer accounts. The attacks, which occurred between February and October 2024, resulted in the theft of data from at least 165 organizations, including AT&T and Ticketmaster, impacting over 100 million individuals. Moucka and co-conspirators exploited accounts lacking multi-factor authentication, using credentials stolen by infostealer malware, and obtained over $2.5 million through extortion and data sales.
Azure Cosmos DB Vulnerability "CosmosEscape" Allowed Access to All Databases
Wiz Research discovered "CosmosEscape," a critical vulnerability in Azure Cosmos DB's Gremlin API that could have allowed attackers to compromise all databases within the service, including Microsoft's internal databases. The flaw enabled attackers to acquire a "Cosmos Master Key" for full read and write access. Microsoft has fully remediated the issue, eliminating the platform-wide key and adding new guardrails.
Coldcard Wallet Flaw Leads to Over $88 Million Bitcoin Theft; Phishing Campaign Emerges
A firmware vulnerability in Coldcard hardware wallets, stemming from a March 2021 integration error that routed seed generation to a deterministic software pseudorandom number generator, has resulted in the theft of at least 1,367.05 BTC, valued at over $88.6 million, from 4,585 addresses. Coinkite, the manufacturer, has released emergency firmware updates and destroyed remaining inventory, while a new phishing campaign is exploiting the situation to install remote access software.
Meta's Muse Image AI Faces Privacy Backlash, Feature Disabled
Meta launched Muse Image, an AI tool allowing users to create images from public Instagram photos, leading to privacy concerns. Following user backlash, Meta disabled the feature, which had allowed images to be generated using others' Instagram posts without their consent.
Gunra Ransomware Exploits Fortinet and Schneider Electric Flaws, Linked to Lazarus Group
The Gunra ransomware group is exploiting vulnerabilities in Fortinet firewall products and Schneider Electric PowerLogic P5 appliances to target critical infrastructure globally. South Korean agencies also warn that North Korea's Lazarus Group is sharing tools and infrastructure with Gunra, with both groups exploiting vulnerabilities in mandatory Korean financial security software.
Critical Rails Active Storage Flaw Allows Arbitrary File Read, Potential RCE
Ruby on Rails has patched a critical vulnerability, CVE-2026-66066, in its Active Storage framework that allows unauthenticated attackers to read arbitrary files from application servers. This flaw, with a CVSS score of 9.5, affects applications using libvips for image processing and accepting untrusted image uploads, potentially exposing sensitive data and leading to remote code execution.
Google Introduces Selfie Video for Account Sign-in and Recovery
Google has launched a new selfie video option for account sign-in and recovery, allowing users to regain access by recording a short video of their face. This feature provides an alternative verification method, particularly useful when traditional authentication methods are unavailable, by comparing a live video to a securely stored reference video.
Teens sentenced to 5.5 years for £29M Transport for London cyber attack
Owen Flowers and Thalha Jubair were sentenced to 5.5 years for a 2024 cyberattack on TfL that caused £29 million in damages. The attack severely disrupted services and breached data of millions. Authorities cite this case as a major enforcement action against young cybercriminals.
Researchers Reveal Security Flaws in AI Coding Agents and Open-Source Mobile Frameworks
Researchers from Hong Kong University have highlighted vulnerabilities in AI coding agents, notably OpenAI Codex and Claude Code, which can be bypassed using techniques like SKILLCLOAK. These techniques allow malicious AI add-ons and agents to evade current security scanners. These findings underscore the need for improved security measures in AI agent marketplaces and software, as current defenses are inadequate.
Dependabot introduces default three-day cooldown for version updates
Dependabot now includes a default three-day cooldown before opening version update pull requests. This change aims to reduce the risk of merging compromised versions immediately after their release, enhancing supply chain security for developers.
Cybersecurity Expert Sentenced for Role in BlackCat Ransomware Scams
Angelo Martino, a former ransomware negotiator, has been sentenced to 70 months for aiding the BlackCat ransomware gang. Collaborating with accomplices, he shared confidential negotiation details, causing victims to lose over $75 million. This highlights vulnerabilities within cybersecurity industries.
Zoom Patches Critical Vulnerability Allowing Device Takeover During Screen Sharing
Zoom has patched critical vulnerabilities in its video conferencing platform that allowed attackers to remotely execute code and take over devices during screen-sharing sessions. Discovered by A Security using fewer than 20 AI prompts, the flaw affected all supported operating systems and required no victim interaction, highlighting the increasing accessibility of advanced exploit development.
New "Pass-ta-key" Attacks Bypass Passkey Protections in Google Password Manager
Researchers from Palo Alto Networks' Unit 42 have identified three "Pass-ta-key" attack methods that allow malware on compromised Windows machines to bypass passkey protections in Chrome's Google Password Manager. These attacks exploit how Chrome stores device keys and re-enrolls devices, enabling silent authentication, installation of attacker-controlled keys, or extraction of synced passkey private keys, demonstrating vulnerabilities in passkey implementations when an endpoint is already compromised.
WebKit Flaws Expose Real IP Addresses for iCloud Private Relay and Proxy Browser Users
Security researchers Talal Haj Bakry and Tommy Mysk discovered three WebKit features that bypass proxy configurations, leading to IP and DNS leaks. These vulnerabilities affect Apple's iCloud Private Relay and other proxy browsers on iOS and macOS, potentially exposing users' real IP addresses and DNS servers. A class action lawsuit has been filed against Apple regarding the iCloud Private Relay flaw.
Jscrambler npm Package Supply Chain Attack Deploys Infostealer
The npm package Jscrambler version 8.14.0 was compromised, executing an infostealer on installation and affecting multiple subsequent versions. Released on July 11, 2026, the package was downloaded nearly 1,500 times before removal. The incident, attributed to credential compromise, highlights security risks in open-source dependencies.
HalluSquatting Attack Exploits AI Hallucinations to Form Botnets
The "HalluSquatting" attack exploits AI hallucinations to inject malicious commands into coding assistants, potentially creating botnets. Researchers from Tel Aviv University and other institutions demonstrated that attackers can pre-register fictitious software names generated by AI. AI models' tendency to hallucinate and act on fake package names can expose systems to widespread malware deployment.
Microsoft Launches MAI-Cyber-1-Flash and Perception for AI Cybersecurity
Microsoft introduced MAI-Cyber-1-Flash, its first AI model specialized in cybersecurity, and Project Perception, an agentic security platform. These tools are designed to identify and remediate software vulnerabilities, with MAI-Cyber-1-Flash integrated into Microsoft's MDASH harness. The company claims the new offerings outperform competitor models on benchmarks and reduce operational costs.
Apple Fixes iCloud+ 'Hide My Email' Vulnerability After Public Scrutiny
Apple has patched a vulnerability in its Hide My Email feature that exposed real email addresses, following public and legal pressure. The issue persisted for over a year despite early warnings from security researcher Tyler Murphy. The flaw raised significant privacy concerns for users.
WordPress wp2shell Vulnerability Exploited; Urgent Patches Released
Two critical WordPress vulnerabilities, dubbed 'wp2shell' (CVE-2026-60137 and CVE-2026-63030), allow unauthenticated attackers to execute code. Affecting versions 6.9.0-6.9.4 and 7.0.0-7.0.1, fixes were released in versions 6.9.5 and 7.0.2. The vulnerabilities, actively exploited, prompted immediate patching, affecting over 500 million sites. WordPress initiated forced automatic updates, while Cloudflare deployed protective measures.
PamStealer Malware Targets macOS for Credential Theft Using Apple's PAM
Researchers have discovered PamStealer, a macOS malware that uses Apple's PAM interface to steal user credentials. This sophisticated malware employs a two-stage delivery system, disguising as the clipboard manager Maccy and utilising stealthy JavaScript for Automation. It highlights emerging threats in macOS security exploiting native Apple frameworks for credential theft.
CERT/CC Reports Hidden Backdoor in Tenda Router Firmware Allowing Admin Access
A vulnerability in various Tenda router firmware versions, CVE-2026-11405, allows unauthorized administrative access via an undocumented backdoor. This flaw poses significant security risks for users as attackers can bypass authentication to control devices remotely. Despite warnings, Tenda has not addressed the issue, leaving affected devices unpatched.
Critical Vulnerabilities in Paperclip AI Platform Allow Remote Code Execution
Two critical security flaws in Paperclip, an open-source control plane for AI agents, could allow attackers to execute commands on network servers or developer computers. The most severe vulnerability, CVE-2026-41679, has a CVSS score of 10.0 and allows remote code execution without prior authentication, while another flaw (GHSA-x8hx-rhr2-9rf7) enables execution on developer machines. A third flaw exposed sensitive data through unauthenticated API routes.
RefluXFS Linux Kernel Flaw Allows Local Root Access on XFS Filesystems
A nine-year-old Linux kernel flaw, dubbed RefluXFS (CVE-2026-64600), allows unprivileged local users to overwrite root-owned files on XFS filesystems and gain persistent root access. This race condition vulnerability affects systems running Linux kernel v4.11 or later with XFS filesystems created with `reflink=1`, including default installations of Red Hat Enterprise Linux, Fedora Server, and Amazon Linux.
FBI and CISA Warn of Russian Phishing Attacks on Signal and WhatsApp Accounts
The FBI and CISA have issued an updated warning about Russian intelligence phishing campaigns targeting Signal and WhatsApp accounts. Attackers are using Signal Backup Recovery Keys to hijack accounts, and the U.S. is offering a $10 million reward for information on the group responsible. The campaign has compromised thousands of accounts of high-profile targets, including government officials and journalists.
Zimbra Releases Critical Security Patches for Classic Web Client
Zimbra has released version 10.1.19 to patch a critical stored XSS vulnerability in its Classic Web Client, which could allow attackers to execute malicious code via crafted emails. Additionally, Zimbra version 10.1.20 addresses multiple vulnerabilities, including command injection and mail forwarding bypass. The updates are crucial to maintain security for users of the Zimbra Collaboration Suite.
Google Launches Gemini 3.5 Flash Cyber for Efficient Vulnerability Management
Google DeepMind has launched Gemini 3.5 Flash Cyber, a cybersecurity AI model to detect and patch vulnerabilities efficiently. Initially available to governments and trusted partners via CodeMender, this model provides a cost-effective alternative to larger AI security models like Anthropic's Mythos. Its introduction is part of a broader release including Gemini 3.6 Flash and 3.5 Flash-Lite, demonstrating Google's continued commitment to cybersecurity innovation.
WhatsApp Launches Username Feature for Enhanced User Privacy
WhatsApp has initiated a rollout of a username feature, allowing users to interact via usernames instead of phone numbers. This move is targeted at enhancing privacy for the platform's three billion users. However, there are concerns about potential impersonation risks, especially in large user markets like India.
EU Mandates Google to Provide AI Rivals Access to Android and Search Data
The European Union requires Google to provide rival AI applications access to Android functionalities and to share anonymized search data under the Digital Markets Act. This aims to prevent Google from leveraging its OS dominance to limit competition. Google cited potential privacy and security concerns with these changes.