Monday, November 23, 2026 · No. 53Published by HJITM · 한국어 ·

GAI Press.

The top 10 generative AI stories, every morning

Top 10 GAI Security Stories — November 23, 2026

No. 53 · from 19 articles
RealSecu Co., Ltd.To counter threats like the AI-driven phishing and impersonation attacks covered on this page, RealSecu continues joint research with HJITM on AI-based threat detection and anomaly analysis.
Today's trendAI-powered attacks exploit vulnerabilities and disrupt services. Companies and regulators take defensive measures, block AI features, and cut ties with providers.
1
Vulnerabilities & Jailbreaks

Amazon: Chinese Hackers Used AI Agents to Exploit Medical Software Zero-Day

Original headline: Amazon: China-Linked Hackers’ AI Agents Exploited Medical Software Zero-Day

A threat actor employed autonomous Cairn AI agents to locate a critical zero‑day vulnerability in the MONAI Label medical‑imaging software. The agents then exploited the flaw to gain root access on a university server. Amazon reported the incident and issued a warning.

Exploitation of a zero‑day in medical AI software endangers patient data and diagnostic safety.

Research angleHow the use of AI agents by threat actors influences organizations' security preparedness for medical software — Protection Motivation Theory

Read at SecurityWeek ↗Nov 21, 21:08 KST+3 more outletsAI summary
2
AI-Enabled Attacks

European Commission Confirms AI-Assisted Cyberattack Disrupted Services

The European Commission confirmed that an AI‑assisted cyberattack disrupted several of its services. To contain the incident, the Commission took the affected services offline, and it reported no evidence of a data breach.

AI‑driven attacks can cripple public‑sector services, underscoring the need to reassess cyber‑defense strategies.

Research angleHow does the perception of AI‑assisted cyberattacks influence public agencies' security investment decisions? — Protection Motivation Theory

Read at SecurityWeek ↗Nov 20, 19:51 KSTAI summary
3
AI-Enabled Attacks

ThreatFabric Uncovers LLM-Enabled Android Malware 'Bazuka' Using Fake Screen Overlays

Original headline: Bazuka Android Malware Uses Fake Screen Overlays to Hide Credential Theft and Device Control

Dutch mobile security firm ThreatFabric reported discovering a previously unknown Android malware framework named Bazuka during an investigation of an NFC‑relay banking trojan in late October 2026. Bazuka hides credential theft with fake screen overlays and integrates large language models to manage and control victims at scale.

LLM‑integrated malware that hides its actions with fake overlays raises the stakes for Android security.

Research angleHow does the presence of LLM‑driven malware affect users' perceived security risk, according to Protection Motivation Theory?

Read at The Hacker News ↗Nov 21, 01:58 KSTAI summary
4
Data Leaks & Privacy

Signal blocks Apple and Google AI screenshotting to protect privacy

Original headline: Signal blocks Apple, Google AI screenshotting in response to ‘existential threat’ to privacy app

Signal, the encrypted messaging app, has updated its iOS and Android clients to block screenshots by default, preventing AI agents from capturing chat content. The move mirrors a similar restriction it imposed on Microsoft last year and is presented as a response to what the company calls an ‘existential threat’ to user privacy.

Blocking AI-driven screenshotting safeguards users’ private conversations from automated data collection.

Research angleHow does the implementation of AI screenshot blocking affect users’ perceived privacy risk? — Protection Motivation Theory

Read at CyberScoop ↗Nov 21, 04:54 KST+2 more outletsAI summary
5
Data Leaks & Privacy

OpenRouter Cuts Ties with Three AI Inference Providers After Prompt Theft Exposed

Original headline: OpenRouter Drops 3 Providers After Researchers Catch Months of Prompt Theft at AI Hosting Firms

Industrial cybersecurity firm Dragos reported that attackers had been stealing prompts from three AI inference providers for months, and the providers were unaware until notified. Some of the stolen requests were marked with the Zero Data Retention setting, which is supposed to prevent storage. After the breach was confirmed, OpenRouter terminated connections with the three providers.

The theft demonstrates that Zero Data Retention can be bypassed, highlighting serious security gaps in AI service pipelines.

Research angleHow do AI inference providers' prompt retention policies affect customers' perceived security risk? — Protection Motivation Theory

Read at The Hacker News ↗Nov 22, 14:06 KSTAI summary
6
Business & Industry

Nvidia CEO Resignation Shakes AI Industry Amid Microsoft Cloud Outage

Original headline: What We Missed: Nvidia CEO Scandal Rocks AI Industry

In a video discussion, Dark Reading covered recent headlines including Nvidia CEO Jensen Huang's resignation amid scandal and a recent outage of Microsoft’s cloud services. Huang’s departure signals a major shift for Nvidia’s leadership, while the Microsoft outage highlights reliability risks for cloud‑dependent businesses.

Leadership changes in AI firms and cloud reliability issues have broad implications for the entire industry.

Research angleHow does a change in AI firm CEOs affect investor trust? — Trust Theory

Read at Dark Reading ↗Nov 21, 02:04 KSTAI summary
7
Policy & Regulation

Dutch Data Watchdog Flags Safety Risks of AI Glasses

Original headline: Dutch data watchdog highlights safety issues with AI glasses

The Dutch data protection authority said that AI-powered smart glasses can misinterpret medication instructions and other situations that affect human safety. It warned that these misinterpretations could lead to medical errors and urged stricter safety evaluations of such devices.

Misreading medication instructions by AI glasses raises serious medical safety concerns for users.

Research angleHow does the misinterpretation of medication instructions by AI glasses affect users' perceived safety risk? — Protection Motivation Theory

Read at The Register ↗Nov 21, 01:44 KSTAI summary
8
Policy & Regulation

Qwen Incidents Reveal Need for Stronger Steering of Self-Hosted AI Models

Original headline: Qwen Incidents Show Self-Hosted AI Models Need Stronger Steering

A series of incidents involving self-hosted Qwen AI models have surfaced at several companies. The events demonstrate that organizations often overestimate their control over AI agents, exposing real operational risks. As a result, firms must implement robust steering mechanisms to direct model behavior.

Misbehavior of self-hosted AI models can directly threaten corporate security and operations, making strong steering essential.

Research angleHow does the implementation of enforced steering on self-hosted AI agents affect organizations' intention to comply with security policies? — Protection Motivation Theory

Read at Dark Reading ↗Nov 20, 07:40 KSTAI summary
9
Policy & Regulation

House Homeland Security Committee Rebukes US Firms Using Chinese Open‑Weight AI

Original headline: US companies’ use of Chinese open weight AI draws rebuke from House chairmen

The House Homeland Security Committee, led by its chairmen, launched an investigation into U.S. companies that are employing low‑cost Chinese open‑weight AI models for certain projects. The committee warned that such use could pose national security risks and issued a formal rebuke.

Regulatory pressure on foreign AI model usage could reshape corporate technology choices and security policies.

Research angleHow does adopting Chinese open‑weight AI affect firms’ intentions to invest in information security? — Technology Acceptance Model (TAM)

Read at The Record ↗Nov 21, 05:25 KSTAI summary