On May 18, 2026, the Dutch FIOD seized over 800 servers and arrested key individuals linked to THE.Hosting, a bulletproof hosting provider associated with EU influence operations. However, telemetry data from the firm ELLIO indicates that automated scanning activity from the network’s address space continued unabated in the days following the law enforcement operation. This persistence indicates a shift in decentralized threat networks: physical hardware seizures are increasingly insufficient when networks can re-provision infrastructure in new jurisdictions, provided their Border Gateway Protocol (BGP) routing announcements remain active.
The persistence of these networks aligns with a broader acceleration in threat cycles. A 2026 biennial report from Truesec shows that the average time for a threat actor to achieve unauthorized access to an organization has decreased from 53 days in 2024 to 2.4 days in 2026, largely due to AI-enabled automation. Despite this acceleration, Nordic CISOs report stable severe incident rates. This resilience stems from improved managed detection and response (MDR) partnerships and a strategic shift toward safeguarding critical business processes rather than relying solely on traditional perimeter defense.
The speed of unauthorized access is further compressed by AI-assisted vulnerability validation. Analysis of nearly 70,000 vulnerabilities shows the time required to develop a functional proof-of-concept (PoC) after a patch release has dropped from 125 days to half a day. This compression creates a visibility gap for organizations relying exclusively on traditional vulnerability scanners. Data shows that 83.2% of critical vulnerabilities experience this gap, where PoCs circulate well before scanning vendors release detection signatures. In response, security teams are moving toward continuous software inventory analysis and Software Bill of Materials (SBOM) correlation to identify affected assets immediately upon disclosure.
While some groups rely on automation, others find success by reverting to physical and social engineering tactics. The Silent Ransom Group (SRG), also known as Luna Moth, has escalated campaigns against law firms by transitioning from remote vishing to physical facility access. When remote social engineering fails to secure access via tools like WinSCP or Rclone, SRG deploys individuals to physical office locations. Posing as internal IT staff, they claim a need to "image" a workstation or perform a backup to resolve a security issue. Once granted physical access, they use external storage devices to exfiltrate data directly, bypassing traditional network boundaries and endpoint detection.
A similar blend of specialized tooling and localized social engineering is targeting Latin America. The BTMOB Android remote access Trojan (RAT) is currently distributed via a malware-as-a-service (MaaS) model for a $5,000 lifetime fee. This provides low-level actors with an APK builder to create unauthorized banking applications. These campaigns focus on government agencies and public-administration systems in Uruguay, Mexico, and Colombia. Threat actors in these regions increasingly favor pure extortion models, bypassing encryption to focus on high-volume data exfiltration. They use regulatory compliance as leverage, threatening to release sensitive citizen data to trigger government fines and public pressure.
As organizations integrate agentic AI, they risk inadvertently reintroducing legacy vulnerabilities into modern systems. Security researchers note that the most significant risks to AI agents are often not flaws in the models themselves, but the absence of traditional security controls. Agentic AI systems are split into deterministic halves (traditional software) and non-deterministic halves (the large language model). Vulnerabilities—such as those recently evaluated in Salesforce and ServiceNow. Often stem from missing input sanitization or authentication gaps where these two halves connect. For example, an AI agent might expose sensitive data simply because it lacks the identity-based access controls normally applied to a human user or a standard API.
Navigating the current environment requires infrastructure-level and identity-centric security. Mitigating bulletproof hosting networks like THE.Hosting relies on coordinated cross-jurisdictional efforts to "blackhole" offending BGP address spaces, rather than relying strictly on physical seizures. For mobile environments, protecting against threats like BTMOB requires strict enforcement of verified app repositories and managing mobile devices with the same rigor applied to workstations. To address the 0.5-day vulnerability validation cycle, we recommend building parallel detection paths that correlate SBOMs against new CVE disclosures the moment they publish, rather than waiting for automated scanners to flag the issue.
While the methodologies—AI agents, BGP routing, and MaaS builders, continue to evolve, the underlying failure points remain familiar. Whether addressing a physical access attempt by the Silent Ransom Group or an authentication bypass in an AI chatbot, organizations can protect their environments through the rigorous application of established cybersecurity principles: identity verification, least-privilege access, and continuous asset visibility. As threat cycles compress, organizations that translate technical exposure into business-process risk will be best positioned to prioritize resources and safeguard their core operations.