Security Affairs AI-cybersecurity newsletter round 7 compiles AI threat items, including malicious Claude Code extension in Cursor marketplace
Security Affairs published round 7 of its weekly AI-cybersecurity newsletter, a curated roundup of research and reporting at the intersection of artificial intelligence and security rather than coverage of a single incident. Items listed include a reported case in which an Accenture contractor connected an AI agent to a Pentagon secure network and sent data to a high-risk website, a security notice about malicious versions of the Anthropic.claude-code extension in the Cursor VS Code Extension Marketplace, the VOLTA infostealer using model-driven victim prioritization, the rise of LLMJacking monetizing compromised cloud and model resources, a "Black October" for AI developer platforms, infrastructure mapping of ARTEX behind South Korean bank breaches, prompt adware hidden in jailbreak bundles, the WorkOS October 30, 2026 data security incident, and a Claude Dispatch flaw enabling browser-extension RCE on the host OS. The newsletter names no CVEs, indicators or attribution of its own; each item links to a third-party source. Operational value is aggregation rather than new technical detail.
Detalhes técnicos
On the Cursor item, the vendor's own notice states that Anthropic.claude-code versions 2.1.312 and 2.1.313 for Windows and macOS were malicious and were not developed by Anthropic, and that the maliciously modified extension was available for roughly eleven hours and ran code attempting to install a persistent SSH backdoor (cursor.com). Outside unusual cases where Cursor was started with administrator privileges, errors in the malicious code caused that step to fail (cursor.com).
Remediação
Users of the Cursor marketplace should verify the installed Anthropic.claude-code extension version and remove 2.1.312 and 2.1.313; Cursor removed the malicious version from its marketplace upon detection and blocked further distribution (cursor.com). For the other items, defenders should consult the primary sources linked by the newsletter.
Linha do tempo
-
Security Affairs AI-cybersecurity newsletter round 7 compiles AI threat items, including malicious Claude Code extension in Cursor marketplace
ler a análise desta fonte ocultar
Security Affairs published round 7 of its weekly AI-cybersecurity newsletter, a curated roundup of research and reporting at the intersection of artificial intelligence and security rather than coverage of a single incident. Items listed include a reported case in which an Accenture contractor connected an AI agent to a Pentagon secure network and sent data to a high-risk website, a security notice about malicious versions of the Anthropic.claude-code extension in the Cursor VS Code Extension Marketplace, the VOLTA infostealer using model-driven victim prioritization, the rise of LLMJacking monetizing compromised cloud and model resources, a "Black October" for AI developer platforms, infrastructure mapping of ARTEX behind South Korean bank breaches, prompt adware hidden in jailbreak bundles, the WorkOS October 30, 2026 data security incident, and a Claude Dispatch flaw enabling browser-extension RCE on the host OS. The newsletter names no CVEs, indicators or attribution of its own; each item links to a third-party source. Operational value is aggregation rather than new technical detail.
Technical Details
On the Cursor item, the vendor's own notice states that Anthropic.claude-code versions 2.1.312 and 2.1.313 for Windows and macOS were malicious and were not developed by Anthropic, and that the maliciously modified extension was available for roughly eleven hours and ran code attempting to install a persistent SSH backdoor (cursor.com). Outside unusual cases where Cursor was started with administrator privileges, errors in the malicious code caused that step to fail (cursor.com).
Remediation
Users of the Cursor marketplace should verify the installed Anthropic.claude-code extension version and remove 2.1.312 and 2.1.313; Cursor removed the malicious version from its marketplace upon detection and blocked further distribution (cursor.com). For the other items, defenders should consult the primary sources linked by the newsletter.