TL;DR
ChatGPT content-sharing features abused for malware delivery; Marimo CVE-2026-39987 exploited with LLM-assisted post-compromise activity; Dutch authorities disrupt numerous-device botnet and arrest two hosting company executives aiding Russian cyberattacks.
Executive Summary
- ChatGPT's share link feature is being weaponized to host fake outage pages delivering malware disguised as the ChatGPT desktop application.
- Threat actors have begun using large language model agents for post-exploitation activity after CVE-2026-39987 exploitation, indicating a new class of automated compromise tools.
- Dutch law enforcement seized 800 servers and arrested two hosting company co-owners who provided infrastructure for Russian state-sponsored cyberattacks, disinformation, and influence operations targeting the EU.
- A new threat actor dubbed GREYVIBE, assessed as Russian-speaking, has been conducting persistent AI-powered cyberattacks against Ukraine since August 2025.
- Malicious NuGet packages impersonating Sicoob (Brazil's largest cooperative financial system) are stealing banking credentials and PFX certificates.
Top Threats Today
1. ChatGPT Share Links Weaponized for Malware Distribution
Severity: HIGH Affected: Technology
Threat actors are abusing ChatGPT's content-sharing feature to display fake OpenAI outage pages that direct users to download malware disguised as the ChatGPT desktop application [1]. This technique exploits user trust in the ChatGPT platform to facilitate credential theft and endpoint compromise.
Sources:[1] BleepingComputer
Recommended Action
- Alert end users to verify ChatGPT status through official OpenAI channels (status.openai.com) rather than links in messages or search results
- Block or monitor suspicious ChatGPT share links (share.openai.com) in email gateways and web proxies
- Enforce application whitelisting or code-signing verification for desktop client downloads
2. LLM Agents Enable Post-Compromise Automation After Marimo Exploitation
Severity: HIGH Affected: Technology
An unknown threat actor has exploited CVE-2026-39987 in a publicly-accessible Marimo instance and subsequently used a large language model agent to conduct post-compromise actions [1]. This represents the first observed use of LLM-powered automation for post-exploitation, indicating attackers are adopting AI-driven lateral movement and data exfiltration ⚠ techniques.
Sources:[1] The Hacker News
Recommended Action
- Audit and patch all publicly-exposed Marimo instances immediately
- Monitor for unusual LLM API calls or autonomous agent behavior in cloud logs (e.g., repeated API queries without human initiation)
- Implement network segmentation to limit lateral movement capability from compromised development tools
3. Dutch Law Enforcement Disrupts Russian Cyberattack Infrastructure, Arrests Hosting Executives
Severity: HIGH Affected: Government, Technology
Dutch authorities arrested the co-owners of two related Internet hosting companies for operating IT infrastructure used by Russia to conduct cyberattacks, influence ⚠ operations, and disinformation campaigns inside the European Union, and seized 800 servers at a local provider [1]. This disruption represents a significant blow to Russian state-backed cyber operations targeting EU entities.
Sources:[1] Krebs on Security
Recommended Action
- Review threat intelligence bulletins from Dutch NCSC and EU agencies for IOCs associated with the seized infrastructure
- Cross-reference organizational logs against known Russian-hosted command-and-control domains and IP ranges from the seized server pool
- Increase monitoring for disinformation and influence campaign indicators originating from hosting providers linked to Russian operations
4. GREYVIBE — Russian-Linked AI-Powered Cyber Campaign Against Ukraine
Severity: HIGH Affected: Government
A previously undocumented threat actor dubbed GREYVIBE, assessed by WithSecure to be a Russian-speaking group operating in the Russian time zone, has been conducting persistent AI-powered cyberattacks targeting Ukraine and Ukraine-related entities since at least August 2025 [1]. The incorporation of AI capabilities into targeting and attack execution suggests sophisticated state-level adversary maturation.
Sources:[1] The Hacker News
Recommended Action
- Coordinate with CISA and Ukrainian SSSCIP for updated GREYVIBE IOCs and TTPs
- Monitor for AI-generated phishing content and socially-engineered prompts targeting Ukrainian government and critical infrastructure staff
- Increase email authentication (SPF, DKIM, DMARC) and anomalous outbound connection alerting
5. Malicious Sicoob NuGet Package Steals Brazilian Banking Credentials
Severity: HIGH Affected: Finance
Researchers discovered a malicious NuGet package masquerading as a C# SDK for Sicoob, one of Brazil's largest cooperative financial systems; versions 2.0.0 through 2.0.4 of “Sicoob.Sdk” siphon client IDs and PFX certificates [1]. This supply-chain attack directly threatens the confidentiality of banking credentials across Sicoob's cooperative member base.
Sources:[1] The Hacker News
Recommended Action
- Audit all NuGet package dependencies in development pipelines for versions 2.0.0 through 2.0.4 of Sicoob.Sdk
- Revoke any PFX certificates or API credentials that may have been compromised
- Implement NuGet package pinning and integrity verification (checksums, signed packages) in CI/CD pipelines
Today's Action Checklist
- ☐ URGENT: Alert users to verify ChatGPT status through official channels; block suspicious share.openai.com links in email gateways
- ☐ Patch all publicly-exposed Marimo instances and review instance logs for CVE-2026-39987 exploitation indicators
- ☐ Cross-reference organizational logs against IOCs from the Dutch law enforcement operation (800 seized Russian-linked servers)
- ☐ Audit NuGet dependencies for malicious Sicoob.Sdk versions 2.0.0–2.0.4 and revoke compromised banking credentials
- ☐ Enable anomaly detection on cloud and email logs for AI-generated attack patterns consistent with GREYVIBE operations