State Hackers, AI Weapons, and Supply Chain Sabotage: Your Weekly Cybersecurity Rundown

The digital battlefield is evolving at a breakneck pace, with adversaries seamlessly blending cutting-edge technology with time-tested tactics. This week’s threat intelligence paints a clear picture: state-sponsored actors are weaponizing generative AI for reconnaissance, sophisticated groups are poisoning open-source software repositories, and attackers are ruthlessly exploiting known vulnerabilities through bulletproof infrastructure. The line between advanced persistent threats and opportunistic crime continues to blur, emphasizing that modern defense requires continuous vigilance, not just against novel zero-days, but against the creative misuse of trusted tools and platforms. Proactive exposure management is no longer a luxury—it’s the critical differentiator.

🔍 State-Backed Hackers Weaponize Google’s Gemini AI for Attacks

Google’s Threat Analysis Group (TAG) has reported a significant escalation in the misuse of public AI tools. The North Korea-linked threat actor UNC2970 was observed using Google’s own Gemini AI model to conduct sophisticated reconnaissance on potential targets. This marks a pivotal moment, demonstrating that generative AI is actively being integrated into the cyber attack lifecycle by advanced persistent threats (APTs). The technology is being leveraged to accelerate research, understand target environments, and potentially draft more convincing social engineering lures.

This activity is part of a broader trend where various hacking groups are exploiting AI for information operations, phishing automation, and even attempting “model extraction” attacks to steal proprietary AI architectures. The use of a widely available, legitimate tool like Gemini highlights a shift towards “living-off-the-land” with AI, making malicious research harder to detect and attribute as it blends with normal user traffic on these platforms.

Why this matters: This isn’t theoretical. State actors are *right now* using the same AI tools available to the public to plan and enhance attacks. It forces a re-evaluation of threat models around AI services and underscores the need for monitoring how these tools are accessed within an enterprise.

Source: Read Source

🐍 Lazarus Group Poisons npm and PyPI with “Graphalgo” Campaign

The Lazarus Group, a notorious North Korean state-sponsored hacking collective, has launched a new software supply chain attack. Dubbed “graphalgo,” this campaign involves planting malicious packages in the npm (JavaScript) and PyPI (Python) repositories. The packages are disguised as legitimate tools, often with names related to algorithms or utilities, and are part of a fake recruitment-themed lure to trick developers into downloading them.

The campaign, assessed to be active since May 2025, demonstrates Lazarus’s continued focus on the open-source ecosystem as a high-value attack vector. By compromising a single developer’s system through a malicious dependency, the group can potentially gain a foothold in countless downstream projects and organizations, enabling espionage and further network penetration.

Why this matters: Supply chain attacks targeting open-source repositories are a critical threat to every organization that uses software. This campaign directly threatens developers and the integrity of the software we all depend on, highlighting the need for strict software composition analysis (SCA) and dependency vetting processes.

Source: Read Source

⚙️ ThreatsDay Bulletin: AI Prompt RCE, Simpler Initial Access, and Quiet Tool Misuse

The latest threat intelligence bulletin reveals a pragmatic shift in attacker behavior. Rather than chasing elusive zero-days, adversaries are “leaning harder on what already works.” This includes the quiet misuse of trusted IT and security tools, exploiting familiar workflows, and targeting overlooked exposures that remain in plain sight. A notable technical finding includes methods for achieving Remote Code Execution (RCE) through AI prompt injection.

Another key trend is the decoupling of initial access from post-compromise activities. Attackers are using simpler, more reliable methods to get in the door (like phishing for credentials), then deploying sophisticated toolkits and techniques once inside the network. This “low and slow” approach prioritizes stealth and persistence over noisy, complex exploits.

Why this matters: Defense strategies must adapt. Focusing solely on blocking advanced exploits is insufficient. Organizations need to harden common entry points, strictly monitor the use of admin tools, and assume that determined attackers will eventually find a way in, shifting focus to rapid detection and response inside the network.

Source: Read Source

📊 The Stark CTEM Divide: Why 84% of Security Programs Are Falling Behind

A 2026 market study reveals a dramatic performance gap linked directly to security framework adoption. Organizations that have implemented a Continuous Threat Exposure Management (CTEM) program demonstrate 50% better attack surface visibility and a 23-point improvement in key security metrics compared to those that haven’t. The study indicates that 84% of security programs are lagging due to the absence of this continuous, process-driven approach.

CTEM moves beyond periodic vulnerability scans to a continuous cycle of scoping, discovery, prioritization, validation, and mobilization. This proactive stance allows organizations to systematically identify and remediate critical exposures before they can be exploited, rather than reacting to breaches after the fact.

Why this matters: Budget and tools alone aren’t creating resilience. The data shows that adopting a structured, continuous framework like CTEM is the single biggest factor separating leading security programs from the rest. It’s a blueprint for turning overwhelming threat data into actionable, prioritized defense.

Source: Read Source

🛡️ 83% of Ivanti EPMM Exploits Sourced from Single Bulletproof Hosting IP

The exploitation of a recent Ivanti Endpoint Manager Mobile (EPMM) vulnerability provides a case study in attacker infrastructure. Threat intelligence firm GreyNoise found that 83% of exploit attempts (346 out of 417 sessions) originated from a single IP address hosted on “PROSPERO,” a known bulletproof hosting service. This service is designed to shield malicious activity from takedown requests.

This concentration shows how threat actors efficiently weaponize new vulnerabilities by deploying scalable attack infrastructure from abuse-tolerant providers. It allows even less sophisticated actors to launch widespread exploitation campaigns rapidly, leveraging the work of initial vulnerability researchers and exploit developers.

Why this matters: It highlights the industrial-scale nature of vulnerability exploitation. Defenders can use this intelligence to block known malicious IP ranges proactively. It also underscores the critical importance of rapid patching—once an exploit is public, automated attacks from infrastructure like this will follow within hours or days.

Source: Read Source

Key Takeaways for Security Teams:

• AI is an Active Attack Tool: State actors are using public LLMs like Gemini for reconnaissance. Monitor and policy the use of generative AI in your environment.
• Audit Your Dependencies: The Lazarus “graphalgo” campaign is actively poisoning npm/PyPI. Implement strict software supply chain security controls.
• Focus on Fundamentals: Attackers are exploiting simple initial access and trusted tools. Harden common entry points and monitor for anomalous use of admin utilities.
• Adopt a Continuous Framework: The data is clear—implementing a CTEM program is the most significant step to improve visibility and reduce risk.
• Patch with Urgency: Exploitation of new vulnerabilities is automated and scaled through bulletproof hosting. Speed from disclosure to remediation is a primary defense metric.