GAI Press.

Malicious Claude Code Extension Delivered via Cursor Supply Chain Attack

Original headline: Malicious Claude Code Extension Served in Cursor Supply Chain Attack

A malicious Claude code extension was distributed through the supply chain of the Cursor development tool. The extension performed host reconnaissance and attempted to establish a reverse SSH tunnel.

Supply‑chain delivery of LLM‑based malicious extensions shows that even trusted developer tools can become vectors for sophisticated attacks.

Research angleHow does the presence of LLM‑based malicious extensions in software supply chains affect developers’ security awareness? — Protection Motivation Theory

Chainalysis Discloses Data Breach After AI-Driven Attack

Original headline: Chainalysis Discloses Data Breach Following AI-Driven Cyberattack

Chainalysis announced that its systems and internal tools were compromised in a cyberattack that leveraged AI. The attackers employed so‑called ‘offensive agentic’ techniques to gain access and exfiltrate customer data, resulting in a data breach.

AI‑driven attacks are raising the scale of data security threats for enterprises.

Research angleHow does perceived risk of AI‑driven attacks influence firms’ security investment decisions? — Technology Acceptance Model (TAM)

400 Chinese‑Developed AI Apps Secretly Offer Sexually Abusive Deepfake Creation

Original headline: Hundreds of Mundane AI Apps Are Hiding Sexually Abusive Deepfake Features

WIRED investigated roughly 400 AI apps and websites linked to Chinese‑language developers. While the services appear as ordinary AI tools, they conceal secret capabilities that let users generate violent, sexually abusive deepfake images and videos.

The ability to mass‑produce sexually abusive deepfakes heightens online victimization and complicates regulatory responses.

Research angleDoes the hidden deepfake functionality in AI apps affect user trust, and can Protection Motivation Theory explain the resulting security behaviors? — Protection Motivation Theory

Researchers Find New VOLTA Infostealer Uses Claude, Gemini, Kimi to Value Victims

Original headline: New VOLTA Infostealer Uses Claude, Gemini, and Kimi AI to Rank Victims by Financial Value

Security researchers have identified a Windows infostealer named VOLTA‑AUTODOXER that leverages large language models to rank stolen data by its monetization potential. The malware harvests passwords and cookies from Chromium‑based browsers as well as data from Outlook profiles and Slack workspaces.

Using LLMs to assess data value enables attackers to prioritize high‑value targets more efficiently.

Research angleHow does LLM‑driven valuation of stolen data affect organizations' security awareness? — Protection Motivation Theory

AITX Highlights SARA's Credentials as Its Agentic AI Security Professional Goes OpenToWork

Original headline: AITX's RAD Puts SARA's Qualifications on Display as Its Agentic AI Security Professional Goes #OpenToWork

AITX introduced its agentic AI security professional, SARA, showcasing its qualifications via the RAD platform. CEO Steve Reinharz discussed the launch in Detroit, inviting the security industry to explore how SARA can contribute to their operations.

The rollout signals that autonomous AI security agents are entering the market, prompting industry scrutiny and adoption.

Research angleHow does the adoption of agentic AI security professionals affect firms' incident response efficiency? — Technology Acceptance Model (TAM)

Is Your SOC Ready for AI Agents? A Maturity Check Before You Deploy

Forty percent of security teams now use AI every day, and another 56% are testing it. Only 4% have no plans to adopt it, according to Prophet Security's State of AI in Security Operations 2026 survey of more than 250 CISOs, SOC leaders and practitioners. For the teams still testing, the next decisi…

Microsoft's AI Risk Report Overlooks Government Agencies' Exposure Through Microsoft Bugs

Original headline: Microsoft's AI Risk Report Misses This Key Detail

The latest Digital Defense Report from Microsoft identified government agencies as prime targets for AI-driven attacks. However, the report failed to note that these agencies are being broadly targeted through vulnerabilities in Microsoft software.

The omission highlights that government agencies face significant AI threats via Microsoft software bugs, a risk not reflected in the report.

Research angleHow do Microsoft software bugs affect government agencies' perceived security of AI systems? — Technology Acceptance Model (TAM)

Token thieves drive companies to over $100K daily illicit LLM charges

Original headline: Token thieves costing companies $100K+ per day in illicit LLM charges

Companies are losing more than $100,000 per day due to token theft that fuels unauthorized LLM usage. Attackers exploit compromised API keys to generate charges, prompting firms to strengthen key‑management and monitoring controls.

Illicit LLM usage imposes massive daily financial losses on enterprises.

Research angleHow does the robustness of an organization’s LLM API key management affect the risk of token theft? — Protection Motivation Theory

Thieves Steal Environment Variables from Trigger AI Workflow Tool

Original headline: Thieves pilfer environment variables from Trigger AI workflow tool

Attackers accessed the Trigger AI workflow platform and exfiltrated its environment variables, which often contain API keys and other credentials. The breach highlights the risk of storing sensitive configuration data in cloud‑based AI services, prompting some organizations to consider moving workloads back on‑premises.

Exposure of environment variables can reveal critical secrets, amplifying the attack surface of AI workflow services.

Research angleHow does the management of environment variables in cloud‑based AI workflow tools affect users' security perception? — Protection Motivation Theory