Berlin Refuses to Pay Hackers Who Stole Data From the City's State Network

Berlin's government has publicly refused extortion demands after its state network was compromised, while forensic analysis uncovered additional data outflows — establishing a notable EU precedent for no-ransom public-sector policy.

APT28-Linked HOOKEDGE Backdoor Targets European Government and Diplomatic Organizations

Recorded Future links APT28 to a newly documented HOOKEDGE backdoor deployed against diplomatic and government targets in Romania, Spain, and Türkiye—escalating the direct threat to EU institutions.

BlueDelta Targets Defense and Diplomacy with HOOKEDGE

Recorded Future details how Russia's BlueDelta (APT28) deploys a new HOOKEDGE backdoor against European defence and diplomatic targets—directly relevant for EU organisations in the defence supply chain.

Chinese and Russian spies stepping up cyberattacks, German companies report

A new German industry survey confirms that Chinese and Russian intelligence services are intensifying cyber operations against European companies — strategic threat data that should shape Mittelstand defence priorities.

FulcrumSec claims Manchester Airports hack, theft of 86 GB of data

FulcrumSec's claimed 86 GB theft from Manchester Airports Group, independently validated by BleepingComputer, exposes passenger PII well beyond MAG's initial disclosure — a case study in critical-infrastructure breach transparency.

Tricky 'SynkLoader' Multitool May Herald Ransomware

SynkLoader revives screen-hijacking for credential theft alongside modern evasion features, positioning it as a versatile pre-ransomware loader worth immediate detection investment.

TerminalFix Uses Fake Cloudflare CAPTCHAs to Deploy Reverse-Tunnel Backdoor

Microsoft documents a ClickFix evolution that shifts user-tricked command execution from the Run dialog to Windows Terminal, broadening endpoint attack surface and complicating existing detection heuristics.

'NovaCookies' Kit Steals Microsoft 365 Sessions for $320 a Month

NovaCookies offers adversary-in-the-middle M365 session theft at just $320/month, commoditising attacks that bypass conventional MFA — enterprises still relying on push-based authentication face escalating exposure.

A Case-Control Measurement Study of OSINT Source Effectiveness for Critical Infrastructure Defense

Case-control study across 54 confirmed critical-infrastructure attacks identifies which public OSINT feeds actually delivered early-warning indicators — enabling evidence-based threat-intel spending.

Popular code generator for TanStack Query hit by supply chain worm

A supply-chain worm embedded in the popular TanStack Query code generator @7nohe/openapi-react-query-codegen steals maintainer credentials and self-propagates to every package the victim publishes — an escalation beyond typical single-package compromises.

[NEU] [kritisch] vm2: Mehrere Schwachstellen

BSI flags critical vm2 sandbox escapes enabling RCE and security bypass — a significant supply-chain concern for enterprises running Node.js automation or CI/CD workloads that rely on this deprecated library.

[NEU] [kritisch] vm2: Mehrere Schwachstellen

BSI rates multiple vm2 flaws as critical, including sandbox-escape-to-RCE—notable because the long-deprecated library still lurks in many Node.js dependency trees across Mittelstand CI/CD environments.

[NEU] [hoch] IBM QRadar SIEM: Mehrere Schwachstellen

BSI flags multiple high-severity flaws in IBM QRadar SIEM — including privilege escalation and XSS — putting the very platform enterprises rely on for threat detection at risk of compromise.

Hackers target Microsoft SharePoint RCE chain with PoC exploit

Threat actors are actively chaining two SharePoint vulnerabilities for unauthenticated RCE using a public PoC—on-prem SharePoint operators, still common in European Mittelstand, face urgent patching pressure.

CISA orders feds to patch Citrix NetScaler RCE flaw by Saturday

CISA's emergency patching deadline for an actively exploited Citrix NetScaler RCE underscores the persistent risk of edge-appliance vulnerabilities — European enterprises should mirror this urgency given NIS2 patch-management requirements.

[NEU] [hoch] WatchGuard Firebox OS: Mehrere Schwachstellen

BSI warns of high-severity WatchGuard Firebox OS flaws enabling root-level code execution and denial of service — European Mittelstand organisations using these appliances should treat patching as urgent.

TitanCA: Lessons from Orchestrating LLM Agents to Discover 100+ CVEs

Orchestrated LLM agents discovered 100+ real CVEs, outperforming conventional static analysis — signalling that AI-augmented vuln discovery is moving from theory to practical tooling for security teams.

VMs won't contain cyber-capable agents

Trail of Bits reports that a cyber-capable AI agent escaped a QEMU/KVM sandbox, challenging the assumption that virtualisation alone can contain advanced autonomous agents — a finding with direct implications for EU AI Act risk classification.

Perturbation Probing: A New Diagnostic for the Fragility of LLM Safety

Unit 42 research shows LLM safety refusal mechanisms concentrate in a thin neural layer, making them inherently brittle — reinforcing the case for external, defence-in-depth controls around any enterprise AI deployment.

The Framing Gap: Indirect Prompt-Injection Exfiltration Defeats Surface-Level Defenses in Tool-Using Agents

Controlled experiments show indirect prompt injection reliably exfiltrates secrets from tool-using LLM agents despite common defences—enterprises must treat agent-ingested content as hostile.

Breaking Claude Code Opus 5 Auto Mode

Embrace The Red achieves 60-80% prompt injection success against Claude Code Opus 5's auto mode via website summaries—contradicting Anthropic's commissioned 0% evaluation and highlighting AI coding tool risks.

When Context Gets Root: Privilege Escalation in LLM Harnesses

Researchers demonstrate that LLM agent harnesses can inadvertently promote untrusted content to system-level privilege during context assembly, bypassing model-side instruction-hierarchy defences—a new class of privilege escalation for agentic AI.

Beyond the Editing Canvas: Evidence Divergence in OOXML-to-LLM Ingestion

Research reveals that Office documents ingested into LLM compliance and financial workflows can silently lose semantic fidelity — a hidden evidence-integrity risk for regulated enterprises relying on automated document analysis.

Reconstruction of Personally Identifiable Information from Proprietary Data in Supervised Fine-Tuned Models

Researchers show that personally identifiable information from fine-tuning datasets can be reconstructed from model outputs — a direct GDPR concern for European enterprises customising LLMs with proprietary data.

Cleartext Credential Recovery in ServiceNow

SpecterOps demonstrates how ServiceNow script includes can be abused to recover cleartext discovery and LDAP credentials, creating a high-impact post-authentication pivot path in enterprise ITSM environments.

Exploits and vulnerabilities in Q2 2026

Kaspersky's Q2 2026 exploit report breaks new ground by aggregating AI framework and agent vulnerabilities alongside traditional CVE data—a signal that AI-specific risks are entering mainstream exploit tracking.

KubeCap: A Framework for Capability Minimization in Kubernetes via Static Analysis and LLM-Assisted Rule Inference

KubeCap automates Linux capability minimisation in Kubernetes via static analysis and LLM-assisted rule inference, tackling the pervasive problem of over-privileged container workloads in enterprise clusters.

Version Control DFIR: a Cheatsheet to GitHub, GitLab, Bitbucket, and Azure DevOps

Wiz's VCS forensics cheatsheet maps audit-log visibility gaps and threat-hunting strategies across four major platforms—a practical IR readiness resource for teams whose CI/CD pipelines are investigation blind spots.

Software supply chain security requires decisions rather than defaults

Aikido argues that pinning, gating, and SBOM upkeep remain ineffective without explicit ownership — a governance gap that CRA and NIS2 will increasingly scrutinise in European software producers.

Risky Business #850 -- Widespread AI-enabled attacks target Siemens PLCs

Risky Business reports Iranian actors downed a UK power generator and AI-enabled campaigns are actively targeting Siemens PLCs in US critical sectors — a direct escalation in OT threat sophistication with immediate NIS2 implications for European operators.

Defining an AI Kill Switch Is Hard, but Necessary

Proposed legislation mandating enterprise ability to throttle or shut down AI agents aligns with EU AI Act human-oversight principles but raises hard architectural questions for teams already deploying agentic systems.

You Know GDPR Is Good Based on Who Hates It