Cyber Digest
A daily roundup of key cybersecurity developments across sectors
Executive Summary
The cybersecurity landscape on July 10, 2026 is dominated by two heavy themes: AI agent security and supply chain risk. Multiple coordinated disclosures — Friendly Fire, GhostApproval, and Sophos's detection-trigger analysis — collectively reveal that AI coding assistants (Claude Code, OpenAI Codex, Cursor, Amazon Q Developer, Windsurf, Augment, Google Antigravity) introduce novel attack surfaces at the toolchain level, not just at the AI model level. Separately, Microsoft dissected the GigaWiper/BLUERABBIT destructive backdoor and patched the RoguePlanet Defender privilege escalation flaw, while CISA continued adding to the KEV catalog and urging Fortinet hardening. On the regulatory front, the EU is taking member states to court over unimplemented NIS2-style cybersecurity law, and Cash App's owner agreed to a $45M settlement over lax security practices. In Australia, the ACSC issued a critical alert over a large-scale CMS exploitation campaign.
Incident Map
IT / Technology 6 stories
Friendly Fire: AI Coding Agents Can Be Tricked Into Running Malicious Code
Researchers at the AI Now Institute published a proof-of-concept attack called "Friendly Fire" showing that AI coding agents (Claude Code, OpenAI Codex) running in autonomous mode can be hijacked to execute attacker code. When instructed to scan untrusted open-source code for vulnerabilities, the agent itself becomes the vector — running the attacker's code on the developer's machine instead of catching the threat. Tested across multiple versions of both tools.
GhostApproval: Symlink Flaws in Six AI Coding Assistants
Wiz disclosed a vulnerability pattern called GhostApproval affecting Amazon Q Developer, Claude Code, Augment, Cursor, Google Antigravity, and Windsurf. A booby-trapped repo uses unverified symlinks so that when the AI assistant asks permission to write one harmless-looking file, the write actually lands on a sensitive system file. Three of six have shipped fixes; others remain exposed.
AI Coding Agents Triggering Endpoint Security Rules Built to Catch Attackers
Sophos published telemetry from a week of June 2026 showing that AI coding agents (Claude Code, Cursor, OpenAI Codex) are setting off behavioral detection rules designed for human intruders. Common agent activities — decrypting browser credentials, reading the Windows credential store, pulling files via system tools, writing to startup folders — registered at 56.2% for credential access on monitored machines, creating significant alert fatigue.
npm 12 Disables Install Scripts by Default to Reduce Supply Chain Risk
GitHub released npm version 12 with a major security default change: install scripts (`allowScripts`), Git dependencies (`--allow-git`), and remote URL dependencies (`--allow-remote`) all default to off. Users must explicitly run `npm approve-scripts` to permit trusted packages. Granular access tokens (GATs) that bypassed 2FA are also being deprecated.
Microsoft Patches RoguePlanet Defender Flaw (CVE-2026-50656)
Microsoft patched a privilege escalation vulnerability in the Microsoft Malware Protection Engine (`mpengine.dll`) tracked as RoguePlanet (CVE-2026-50656, CVSS 7.8). The race condition could be abused to spawn a SYSTEM-level shell. Patched in version 1.1.26060.3008, nearly a month after public disclosure by researcher Chaotic Eclipse.
Dormant GitHub Accounts Systematically Mapping Corporate Orgs
Datadog Security Labs warned of overlapping campaigns using dormant GitHub "ghost" accounts and compromised PATs/OAuth tokens to enumerate corporate GitHub organizations, repositories, and user accounts. Over 50 dormant accounts used alongside dozens of compromised legitimate accounts. In select cases, attackers successfully cloned private repositories.
Defence 2 stories
NSA Revives 'Tailored Access Operations' Name for Elite Hacking Unit
The NSA has revived the "Tailored Access Operations" (TAO) designation for its elite offensive hacking unit. TAO was originally active during the early 2000s and was publicly disclosed by Edward Snowden in 2013. The revival signals a renewed organizational focus on computer network exploitation.
Britain Plans to Build Autonomous AI 'Cyber Shield' to Defend Nation
The UK government unveiled plans to develop an autonomous AI-powered "Cyber Shield" to defend national infrastructure. The initiative aims to automate threat detection and response at machine speed, reducing reliance on human-in-the-loop incident response.
General / Cross-Sector 1 story
Global Operation First Light 2026: ~6,000 Arrests, $293M Seized
A coordinated global anti-fraud operation involving 97 countries resulted in 5,811 arrests and the seizure of $293 million in illicit assets. Codenamed First Light 2026, the operation targeted social engineering scams and associated money laundering from January to April 2026. Over 142,000 victims identified globally.
Financial Services 1 story
Cash App Owner to Pay $45 Million to Settle Lax Security Allegations
Block, Inc. (owner of Cash App) agreed to pay $45 million to settle allegations of inadequate security practices. The settlement addresses claims that the company failed to implement reasonable security measures, leading to user data exposure and financial losses.
Government 3 stories
EU Takes Member States to Court Over Unimplemented Cybersecurity Law
The European Commission is taking multiple member states to court for failing to transpose cybersecurity directives (NIS2) into national law. The legal action underscores growing EU impatience with member-state implementation lag, even as cyber threats escalate across the bloc.
EU Unveils Cyber Plan to Reduce Reliance on Foreign AI Systems
The European Commission published a cyber plan aimed at reducing strategic dependence on foreign AI systems, particularly from the US and China. The plan includes investment in sovereign AI training infrastructure and cybersecurity certification for AI supply chains.
Taiwan Charges Two Businessmen in Chinese Espionage Campaign
Taiwan's Ministry of Justice Investigation Bureau charged two businessmen for their alleged roles in a Chinese espionage campaign targeting Taiwanese government agencies and technology firms. The case marks an escalation in Taiwan's response to cross-strait cyber-enabled intelligence operations.
IT / Technology (Vulnerabilities) 4 stories
GodDamn Ransomware Uses PoisonX Driver to Disable Endpoint Defenses
A new ransomware family called GodDamn leverages the PoisonX kernel driver to neutralize security software as a defense evasion strategy. First spotted in the wild on May 21, 2026. Symantec assesses it as a rebrand of Beast ransomware (itself an evolution of Monster), with the developer tracked as Hyadina. Attackers used AnyDesk for remote access and NirSoft-based credential harvesting.
New GigaWiper Windows Backdoor Bundles Disk Wiping, Fake Ransomware, Spyware
Microsoft analyzed GigaWiper, a destructive Windows backdoor combining three older destructive programs into one: full disk wiping, boot drive overwriting, and fake ransomware that encrypts files with an unsaved key. Same malicious files also identified as BLUERABBIT by Binary Defense. Google Threat Intelligence Group ties the malware to a likely Iran-linked threat actor.
Fake 7-Zip Installers Turn Devices Into Residential Proxy Nodes
Infoblox disclosed a threat actor dubbed Lurking Lizard operating a residential proxy business via trojanized 7-Zip installers on lookalike domains (including `7zip[.]com`). Over 230 lookalike domains; activity dating to at least August 2022. The actor also impersonates major proxy providers and runs fake review sites. IPIDEA's infrastructure was dismantled by Google in January 2026.
CISA Adds Multiple KEVs, Urges Fortinet Hardening
CISA added several new Known Exploited Vulnerabilities to the catalog over the past week, including alerts urging organizations to harden Fortinet devices following reports of credential exposure. Separately, CISA published a joint advisory on Russian intelligence services continuing to target commercial messaging applications.
Government (Australia) 1 story
ACSC Critical Alert: Large-Scale Exploitation Campaign Targeting CMS
The Australian Signals Directorate's ACSC issued a critical-rated alert regarding a large-scale exploitation campaign targeting various vulnerabilities in website content management systems (CMS). The advisory is applicable to small & medium businesses, organisations & critical infrastructure, and government audiences.
Manufacturing 1 story
Latvian Forestry Company Still Restoring Systems Weeks After Ransomware Attack
A Latvian forestry company continues restoring systems weeks after a ransomware attack that severely disrupted operations. The extended recovery timeline illustrates the disproportionate impact of ransomware on smaller industrial firms with limited IT/OT resilience.
Media & Entertainment 1 story
Greek Victims File Lawsuit Against Intellexa Over Predator Spyware
Greek citizens filed a lawsuit against Intellexa, the maker of the Predator spyware, alleging illegal surveillance. The case adds to mounting legal pressure on the commercial spyware industry, following similar actions in other European jurisdictions.
Analytics
Source Reliability Index
| Tier | Label | Description |
|---|---|---|
| ● Tier 1 | Very High | Official / first-party |
| ● Tier 2 | High | Established cyber journalism |
| ● Tier 3 | Moderate | General tech/news media |
| ● Tier 4 | Low | Social / unverified |