Wednesday, November 18, 2026 · No. 48Published by HJITM · 한국어 ·

GAI Press.

The top 10 generative AI stories, every morning

Top 10 GAI Security Stories — November 18, 2026

No. 48 · from 46 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 trendToday’s stories show attackers leveraging AI—deepfakes, compromised keys, and AI‑driven exploits to target platforms, hardware, and creative content. Defenders responded with tighter OS permissions, new anti‑deepfake tools, faster bug‑disclosure rules, and government‑focused security offerings.
1
Deepfakes & Fraud

Sam Altman Not Safe from AI Deepfakes on Instagram

Original headline: Not even Sam Altman is safe from AI deepfakes on Instagram

Meta spokesperson Tracy Clayton said that the post by Wang is labeled as AI-generated content. She also noted that despite Altman making his likeness available for AI generation, deepfakes featuring him have appeared on Instagram.

The incident underscores the growing risk of AI-generated impersonations of tech leaders on social media.

Research angleHow do AI-generated deepfakes affect public trust in corporate leaders? — Trust Theory

Read at The Verge ↗Nov 17, 21:59 KSTAI summary
2
Vulnerabilities & Jailbreaks

OpenAI’s GPT-6 Cyber Uncovers 35 Apple Hardware Vulnerabilities

Original headline: OpenAI’s New Hacking Model Found Dozens Of Holes In Apple Tech

Researchers used GPT-6 Cyber to identify 35 vulnerabilities across Apple hardware. Some of these flaws could enable attackers to hijack iPhones and Macs via a malicious website.

The discovery of numerous Apple hardware flaws highlights the growing security risk from AI-driven hacking tools.

Research angleHow does the adoption of AI-based vulnerability detection models affect corporate security policy formation? — Technology Acceptance Model (TAM)

Read at Forbes ↗Nov 17, 20:24 KST+7 more outletsAI summary
3
Policy & Regulation

Apple Tightens macOS App Permissions to Limit Data Access for AI Agents

Original headline: Apple Tightens macOS App Permissions to Limit Data Access Amid AI Agent Security Risks

Apple introduced new permission controls in the macOS 27.3 beta 2 that let users grant more granular data access to apps, including those that employ AI agents. The update enhances transparency and adds selective access options for the Device Control, Data Access, and Full Disk Access APIs.

The measure is a critical step to mitigate data leakage risks posed by the misuse of AI agents.

Research angleHow does the adoption of AI agents affect organizations' intention to comply with data access policies? — Protection Motivation Theory

Read at The Hacker News ↗Nov 17, 20:07 KSTAI summary
4
Defensive Tech

OpenAI Makes Cybersecurity Tool Daybreak Available to NATO and European Governments

Original headline: OpenAI opens its cybersecurity tool Daybreak to NATO and European governments

OpenAI announced that its cybersecurity tool Daybreak is being made available to NATO and several European governments, including Germany, France and Poland. The EU agency ENISA has already used the tool to uncover real vulnerabilities in EU software, and the NATO Communications and Information Agency (NCIA) said it is heavily leveraging the underlying models.

The deployment of an AI‑driven security tool to NATO and EU members strengthens collective cyber‑defense capabilities.

Research angleHow does the adoption of AI security tools affect governmental cyber‑vulnerability detection efficiency? — Technology Acceptance Model (TAM)

Read at The Decoder ↗Nov 18, 00:02 KSTAI summary
5
Defensive Tech

Spotify launches tool to prevent deepfake songs from being linked to real artists

Original headline: Spotify’s new tool aims to stop deepfakes from being attributed to real artists

Spotify has introduced a new tool designed to stop deepfake recordings from being attributed to real artists. This move follows a report by Billboard that Sony Music asked digital platforms to remove more than 260,000 deepfake songs, nearly twice the number disclosed in March.

Blocking deepfake tracks safeguards artists’ rights and maintains trust in streaming services.

Research angleHow does the implementation of deepfake detection tools affect artists’ perceived protection of their intellectual property rights? — Protection Motivation Theory

Read at TechCrunch ↗Nov 17, 23:00 KSTAI summary
6
AI-Enabled Attacks

Google Warns Fully Autonomous AI Attacks Could Appear Within Three Months

Original headline: Fully Autonomous Agentic Attacks Are Three Months Away, Google Warns

Google’s Threat Intelligence Group warned that agentic penetration‑testing frameworks will enable fully autonomous AI attacks within three months, marking a paradigm shift in cyber threats. The agency cited a recent incident where Russian‑speaking hackers carried out the first fully automated AI‑driven cyberattack.

The imminent rise of self‑directed AI attacks could outpace current defensive measures, forcing a rethink of security strategies.

Research angleHow does the adoption of AI‑driven autonomous attack detection affect firms’ security investment efficiency? — Protection Motivation Theory

Read at Infosecurity Magazine ↗Nov 17, 18:30 KST+2 more outletsAI summary
7
Data Leaks & Privacy

Threat Actors Use Compromised Fireworks Key to Breach Cursor via Malicious Claude Extension

Original headline: Cursor Traces Malicious Claude Code Extension to Third-Party Breach

A compromised key issued to AI infrastructure provider Fireworks was leveraged by threat actors to access Cursor’s data and credentials. The breach was facilitated by a malicious Claude code extension.

The incident highlights supply‑chain risks where compromised AI service credentials can expose downstream platforms.

Research angleHow does the management of AI infrastructure authentication keys affect third‑party data breach risk? — Protection Motivation Theory

Read at Dark Reading ↗Nov 18, 03:38 KSTAI summary
8
AI-Enabled Attacks

Amazon: Hackers Use Stolen AI API Keys to Power Agents That Hunt for More

Original headline: Amazon says hackers are using stolen AI API keys to power agents that hunt for more

Amazon said that threat actors have stolen AI API keys and are using them to run AI agents, likely built on the ARTEX platform, which then steal large language model (LLM) keys from exposed AI gateways and search for additional targets. The FBI has issued guidance on how to detect and mitigate such activity.

The abuse of stolen AI API keys to automate credential theft amplifies cloud security risks and forces organizations to reassess defenses.

Research angleHow do corporate AI security policies affect the effectiveness of defenses against hacker attempts to steal API keys? — Protection Motivation Theory

Read at VentureBeat ↗Nov 17, 22:50 KSTAI summary
9
Vulnerabilities & Jailbreaks

Anthropic Shortens Serious Bug Disclosure Deadline to 30 Days

Original headline: Anthropic Cuts Disclosure Deadline to 30 Days For Serious Bugs

Anthropic announced that it will publish any serious flaws discovered by its AI models within 30 days of being reported. The shortened disclosure window aims to accelerate vulnerability sharing and improve security responses across the industry.

A faster bug disclosure timeline helps reduce AI‑related security risks by enabling quicker remediation.

Research angleHow does the length of an AI vendor’s vulnerability disclosure period affect organizational security awareness? — Protection Motivation Theory

Read at Infosecurity Magazine ↗Nov 18, 00:30 KSTAI summary
10
AI-Enabled Attacks

Trigram Campaign Impersonates About 50 AI and Blockchain Brands

Original headline: Trigram Campaign Impersonates About 50 AI and Blockchain Brands to Target Devs with Infostealers

Stormshield reported that a large-scale campaign called Trigram is impersonating roughly 50 AI and blockchain brands to trick developers into running an infostealer themselves. The operation targets developers in Europe and the United States, aiming to harvest sensitive information.

By luring developers to execute malware, the campaign greatly expands the risk of large-scale data theft.

Research angleHow does brand impersonation in AI and blockchain sectors affect developers’ security awareness? — Protection Motivation Theory

Read at The Hacker News ↗Nov 18, 00:17 KSTAI summary