AI Security & Threats
AI-powered attacks, LLM vulnerabilities, and machine learning security
46 articles
The GUARD Act: Congress moves to shield minors from AI companions, but can technology keep up?
A new Senate bill, the GUARD Act, aims to bar minors from AI companions and mandate disclosures. But can technology truly enforce such a digital barri
Zealot shows what AI is capable of in a staged cloud attack
A new AI agent named Zealot, developed by researchers, can autonomously hack cloud environments in minutes, proving AI attacks can outpace human defen
Everyone’s building AI agents. Almost nobody’s ready for what they do to identity
Anthropic's powerful Mythos AI discovered thousands of critical vulnerabilities, highlighting a greater threat: AI agents are poised to dismantle digi
Claude Mythos fears startle Japan's financial services sector
Global financial institutions are panicked over a hypothetical superhacker AI model named "Claude Mythos." Cyber experts explain the reality behind th
Parsing agentic offensive security's existential threat
Frontier LLMs may soon automate entire cyberattacks, sparking a debate: will this lead to cybersecurity annihilation or an unprecedented defensive opp
China's AI hacking claims: Breakthrough or strategic bluff?
Chinese firm 360 Digital Security Group's claim of an AI finding 1,000+ vulnerabilities has been met with skepticism and comparisons to mythical AI.
AI can autonomously hack cloud systems with minimal oversight, researchers find
Palo Alto Networks' Unit 42 demonstrates "Zealot," an AI system that can autonomously execute a full cyberattack on cloud environments.
Pentagon grapples with securing AI as it moves toward autonomous warfare
A senior military leader calls autonomous weapons “essential,” highlighting the Pentagon's immense challenge in securing AI against sophisticated cybe
The dual reality of Iran's political prisoners: Real suffering, AI-generated propaganda
A viral narrative claims Trump saved Iranian women from execution, but the accompanying images are AI deepfakes that exploit a genuine human rights cr
Elon Musk fails to appear for questioning by French police over sexualized AI images on X
An analysis of the French investigation into X over AI-generated child abuse images, and why the non-appearance of its top executives signals a seriou
A hypothetical design flaw in an AI protocol reveals real-world supply chain threats
A deep dive into a hypothetical AI vulnerability that reveals real-world threats to the AI supply chain, from remote code execution to systemic compro
Vulnerability in Cursor AI allowed remote takeover of developer machines
A multi-stage flaw in the Cursor AI IDE could be triggered by opening a malicious file, leading to remote code execution and full system access.
Every old vulnerability is now an AI vulnerability
AI's primary danger isn't creating new bugs, but its power to amplify and accelerate the exploitation of existing, unpatched vulnerabilities.
White House deepens engagement with Anthropic over frontier AI security
A White House meeting with Anthropic's CEO signals a major government push to address frontier AI's unique security and national security risks.
Lawmakers' closed-door AI meetings reveal deep fears of societal destruction
A private meeting between tech titans and U.S. senators exposed profound anxieties over AI's potential for catastrophic risk, moving the debate from t
Ghost breaches: How AI-mediated narratives have become a new threat vector
Three incidents. No actual breaches. Full-scale crisis response. AI hallucinations are creating a new threat vector that most organizations are unprep
OpenAI's new cyber model signals a new front in the AI security arms race
OpenAI's GPT-5.4-Cyber, a model for defenders, enters the field after Anthropic's offensive AI reveal, escalating the AI-driven cybersecurity arms rac
Unverified 'Model Context Protocol' flaw: a theoretical blueprint for AI supply chain attacks
A report on a design flaw in a purported Anthropic protocol remains unverified, but it exposes the theoretical risk of AI models becoming vectors for
Beyond the hype of GPT-5.4-Cyber: How AI is really shaping the future of cyber defense
Speculation about OpenAI's GPT-5.4-Cyber highlights a real trend: AI is escalating the cyber arms race. Here's how it empowers both attackers and defe
OpenAI expands Trusted Access for Cyber program with new GPT 5.4 Cyber model
OpenAI's new GPT 5.4 Cyber model and expanded access program puts it in direct competition with Anthropic, raising questions about control over powerf
‘Mythos-ready’ security: CSA urges CISOs to prepare for accelerated AI threats
The Cloud Security Alliance urges CISOs to prepare for 'Mythos', a future class of AI-driven cyber threats that will shrink attack timelines to minute
US and UK cyber leaders assess threat from advanced AI hacking model
New reports from US and UK security experts reveal the offensive cyber capabilities of a test AI model, signaling a new era of AI-driven threats.
The Mythos incident: When AI closes the gap between detection and disaster
Anthropic's hypothetical 'Mythos' AI autonomously exploited zero-days in all major OSes, highlighting a critical 'post-alert gap' where detection is t
GrafanaGhost exploit bypasses AI guardrails for silent data exfiltration
A new chained exploit, GrafanaGhost, uses AI prompt injection and a URL flaw to silently steal sensitive data from popular Grafana dashboards.