Attackers are actively chaining three JFrog Artifactory vulnerabilities to seize administrative control of self-hosted build infrastructure, while the Dutch NCSC warns that exploitation of two critical Check Point VPN flaws is imminent. These two events, along with Mandiant's documentation of a hijacked AI coding-assistant session spreading a worm across roughly 100 repositories, define a 48-hour threat cycle where software supply chains and perimeter devices are under direct assault.
JFrog Artifactory chain grants full admin takeover
Wiz Research confirmed in-the-wild exploitation between August 15 and September 8, 2026, of CVE-2026-42016, CVE-2026-42018, and CVE-2026-82329. Attackers combined an improper authentication flaw, a token-validation weakness, and an authentication bypass to gain administrative access on self-hosted Artifactory instances. They then deployed persistent rogue admin accounts, malicious Groovy plugins, and Rust-based backdoors to maintain long-term access. Wiz found that 49-62% of organizations remained vulnerable weeks after patches shipped, leaving a substantial exposure window across enterprise build pipelines.
Artifactory sits at the center of software build and release workflows. Administrative compromise gives attackers a direct foothold to tamper with build artifacts across every downstream consumer, making this a high-value supply-chain vector.
Check Point VPN flaws face pre-exploitation warning
The Netherlands' national CERT assessed both the likelihood and impact of exploitation as high for two CVSS 9.8 vulnerabilities in Check Point VPN Security Gateways and Security Management Servers. CVE-2026-85102 involves an improper certificate-validation flaw, and CVE-2026-85103 is a heap-based buffer overflow in ASN.1 certificate parsing. Both allow unauthenticated remote code execution. The NCSC urged organizations to patch before proof-of-concept exploit code becomes public - a genuine advance warning rather than a post-incident analysis.
VPN gateways remain among the most common initial-access vectors for ransomware operators. Organizations still running vulnerable Check Point deployments have a narrow window to act before mass exploitation begins.
AI coding-assistant session hijacked to spread worm
Mandiant detailed an incident at an unnamed SaaS provider where an attacker hijacked a developer's active AI coding-assistant session. The attacker manipulated the assistant into recommending a poisoned PyPI package, then used the compromised session to steal GitHub OAuth tokens and deploy the self-replicating Shai-Hulud worm. The worm exfiltrated secrets and source code across roughly 100 internal repositories. This represents the first concrete case of AI-assistant session hijacking as a software supply-chain vector, distinct from the browser-agent monoculture risk the Cloud Security Alliance has already covered.
CISOs overseeing AI-assisted development should treat live coding-assistant sessions as a privileged access surface. Dependency checksum verification, secrets isolation from IDE extensions, and controlled internal-repository routing are directly actionable controls against this exact attack chain.
EU Cyber Resilience Act reporting platform goes live
ENISA launched the Cyber Resilience Act's Single Reporting Platform on September 11, 2026, the same day binding vulnerability and incident reporting obligations took effect for manufacturers placing digital products on the EU market. Article 16(1) imposes strict deadlines: a 24-hour early warning, a 72-hour detailed assessment, and a 14-day final report. Any organization selling connected products or embedded software into the EU now owns a live regulatory obligation with deadlines measured in hours, not weeks.
This reporting requirement precedes the CRA's broader December 2027 compliance deadline. Incident response playbooks should route CRA-qualifying events through this platform immediately.
Japan ransomware surge hits SMEs with AI-generated scripts
Cisco Talos reported that ransomware incidents in Japan rose 4.7% year-over-year in the first half of 2026, with small and mid-sized enterprises capitalized under JPY 1 billion representing 80% of victims. Talos assessed with medium-to-high confidence that the Qilin ransomware group is using AI-generated Python scripts to accelerate development of its attack tooling, compressing the time attackers need to iterate. Under-resourced SMEs are absorbing a disproportionate share of the impact.
Why this matters for IT and development teams
The JFrog Artifactory chain is a direct threat to build infrastructure. If your team runs self-hosted Artifactory, verify patching status and audit for rogue admin accounts and unexpected Groovy plugins immediately - 49-62% of orgs in Wiz's sample had not applied available fixes weeks after disclosure. For teams using AI coding assistants, the Mandiant incident makes session hygiene a concrete security requirement: isolate secrets from IDE extensions, enforce dependency checksums, and restrict internal repository routing. On the compliance side, any organization shipping connected products to the EU must now route CRA-qualifying incidents through ENISA's platform within 24 hours. Update your incident response playbooks to include that workflow.
Your membership also unlocks:
