• Adobe's maximum-severity ColdFusion flaw, CVE-2026-48282, was exploited within hours of disclosure, prompting a CISA emergency patching order for federal agencies.
• Accenture and KDDI both confirmed major breaches, with Accenture data offered for sale and KDDI notifying over 12 million people.
• Scattered Spider was redefined as a decentralised cybercrime collective, with its subclusters linked to Okta phishing, SIM swapping, and ransomware.
• AI coding assistants faced multiple disclosures, including GhostApproval symlink flaws, SKILLCLOAK scanner bypasses, and image-based prompt injection via GhostCommit.
• Law enforcement actions intensified, with a BlackCat negotiator sentenced, a Ryuk operator pleading guilty, and Operation First Light yielding thousands of arrests.
• Supply chain attacks hit npm packages jscrambler and @injectivelabs/sdk-ts, delivering credential-stealing binaries and cryptocurrency wallet theft.

Adobe disclosed CVE-2026-48282, a maximum-severity path traversal flaw in ColdFusion's Remote Development Services FILEIO handler, enabling unauthenticated remote code execution. KEVIntel honeypots detected exploitation within two hours of public disclosure, and CISA ordered federal agencies to patch by Friday. Shadowserver tracked nearly 800 internet-exposed ColdFusion instances still vulnerable.

Accenture confirmed an isolated security breach after a threat actor claimed to have stolen 35GB of internal data, including source code, Azure access keys, and RSA and SSH keys. The attacker advertised the dataset on PwnForums and shared a screenshot of an apparent private Azure DevOps repository as proof. Accenture said it remediated the breach source with no impact on operations.

Group-IB identified Scattered Spider as a decentralised collective of independent subclusters sharing tactics and communication channels rather than a single organised group. The report links subclusters to Okta phishing, SIM swapping, cryptocurrency theft, and ransomware activity across multiple industries. Attackers used identity provider impersonation, vishing, and stolen employee information to compromise targets.

Wiz identified GhostApproval, symlink vulnerabilities affecting six AI coding assistants including Claude Code, Cursor, and Amazon Q Developer. Malicious repositories could redirect agent file operations outside the intended workspace, in some cases writing to SSH authorised keys files. Cursor and AWS assigned CVEs for the fixed issues, including CVE-2026-50549.

Check Point Research identified Cavern Manticore, an Iran-linked threat actor targeting Israeli government and IT organisations with a modular .NET command-and-control framework. The framework uses Mixed-Mode compilation, DLL sideloading, and modules for reconnaissance and credential operations. Initial access involved abuse of existing Remote Monitoring and Management software already deployed within targeted organisations.

Proofpoint tracked UNK_MassTraction, a suspected China-aligned cluster exploiting Roundcube vulnerabilities against physics and engineering departments at US and Canadian universities since May 2026. The campaign used CVE-2024-42009 to deploy an IceCube credential stealer, then exploited CVE-2025-49113 to install SquareShell webshells or load the VShell backdoor in memory.

Symantec identified GodDamn ransomware as a rebrand of the Beast family, previously evolved from Monster ransomware. Operators used AnyDesk for remote access, a NirSoft-based credential toolkit, and the PoisonX kernel driver to disable endpoint security before deploying ransomware, encrypting files with the .God8Damn extension across at least 10 hosts.

Microsoft Threat Intelligence identified GigaWiper, a Golang backdoor first observed in October 2025 combining disk wiping, fake ransomware, and system sabotage. The malware includes wiping commands derived from FlockWiper and encryption code from Crucio ransomware, using RabbitMQ and Redis for command-and-control communication and generating unsaved keys that prevent file recovery.

SentinelOne reported suspected China- and India-linked espionage groups targeted Pakistani law enforcement organisations between February 2024 and April 2026. Attackers deployed PlugX, ShadowPad, Cobalt Strike, Remcos and AsyncRAT against the Balochistan Police Complaint Management System, hosting malicious implants disguised as portal updates affecting both police personnel and citizens.

ShinyHunters leaked data on 2.3 million people tied to Moody Bible Institute after the college did not meet extortion demands. Separately, a major medical device manufacturer notified nearly 4 million individuals of a breach involving Social Security numbers and health data.

Threat actors exploited CVE-2026-20896, a critical Gitea Docker image vulnerability allowing authentication bypass through a single HTTP header. The flaw affected images up to version 1.26.2, where reverse-proxy trust settings accepted any source IP, letting attackers impersonate users with known usernames roughly 13 days after public disclosure.

Security researcher Hyunwoo Kim disclosed Januscape, CVE-2026-53359, a use-after-free vulnerability in KVM/x86 shadow MMU emulation allowing guest-to-host escape on Intel and AMD systems. The 16-year-old flaw was used as a zero-day exploit in Google's kvmCTF and can trigger host kernel panics or privilege escalation.

Ubiquiti released updates for UniFi Connect, Talk, Access, Protect and UniFi OS after disclosing 25 CVEs, including command injection, SQL injection, and access control flaws. CVE-2026-50746 carried a maximum CVSS score of 10.0 in UniFi Connect, while CVE-2026-50747 scored 9.9 in UniFi Talk.

Nebula Security disclosed GhostLock, CVE-2026-43499, a stack-use-after-free flaw in the Linux kernel's rtmutex subsystem affecting distributions from version 2.6.39 onward. The bug allows unprivileged local attackers to corrupt kernel memory and achieve privilege escalation or container escape, and was fixed in April 2026.

Binarly disclosed six vulnerabilities in U-Boot's FIT Signature Verification affecting code present since 2013. Two flaws could enable arbitrary code execution during boot, while four cause denial-of-service conditions; all were patched in the U-Boot master branch after disclosure to maintainers.

BeyondTrust disclosed four vulnerabilities in its Remote Support and Privileged Remote Access products, including CVE-2026-40138 and CVE-2026-40139, both scoring 9.2. The critical flaws could let unauthenticated attackers bypass access controls, while two additional issues could cause denial of service or unintended resource access.

Researchers from Hong Kong University of Science and Technology developed SKILLCLOAK, hiding malicious AI agent skills through structural obfuscation and self-extracting packing. Testing against eight scanners with 1,613 malicious skills found self-extracting packing bypassed every scanner above 90%, while a runtime auditor detected 97% of attacks.

ASSET Research Group demonstrated GhostCommit, hiding instructions inside a PNG image referenced by a repository configuration file to trigger multimodal coding agents. The technique caused agents to read a repository's .env file and embed secrets into generated code, while CodeRabbit and Bugbot missed the attack because both exclude images from review by default.

Several jscrambler npm releases were compromised via stolen publishing credentials to execute hidden Rust-built native binaries targeting Linux, Windows and macOS. The malware stole browser credentials, cryptocurrency wallet data, and cloud secrets before Jscrambler revoked the credentials and published a clean release.

Group-IB found updated RedHook Android RAT variants abusing Wireless ADB and the Shizuku framework to obtain shell-level privileges without rooting devices. Distributed through fake government and financial websites, the malware supports 53 command-and-control instructions and has expanded from Vietnam into Indonesia.

Elastic Security Labs identified REF6045, a Mexican banking fraud operation delivering SCMBANKER via fake CAPTCHA pages. The PowerShell toolkit monitors banking sessions, captures screenshots, and manipulates clipboard data to redirect funds, with components dating back to at least October 2025.

Former DigitalMint employee Angelo Martino was sentenced to 70 months in prison for participating in BlackCat ransomware attacks against at least five US organisations between 2023 and 2025. Court documents state he shared confidential insurance limits and negotiation details with the operators while acting as a negotiator.

Armenian national Karen Serobovich Vardanyan pleaded guilty to conspiracy and computer fraud charges for deploying Ryuk ransomware against US organisations between November 2019 and April 2020. Vardanyan and co-conspirators received approximately 1,610 bitcoins in ransom payments, valued at more than $15 million.

Spanish police arrested a man in Palencia suspected of supporting CyberArmy of Russia Reborn and Z-Pentest, groups linked to attacks on critical infrastructure in the US and Europe. Investigators said the suspect provided logistical support to a CARR member and participated in activities attributed to NoName057(16).

The NHS launched a staff awareness campaign warning that employees accessing patient records without legitimate reason may face criminal prosecution and dismissal. The guidance follows the dismissal of 11 staff for unlawfully accessing records linked to the 2023 Nottingham attacks.