Every analysis published on PhilipHall.com, newest first. Each piece covers a cyber AI, agent security, automation or enterprise defence development: what happened, why it matters and what organisations can do about it, with sources linked.
Browse by topic: Cyber AI, AI agent security, AI-enabled threats, Breaches, AI governance, Privacy. Read about Philip Hall and the editorial standards.
OpenAI’s AI Agent Hacked Hugging Face. Why Your Sandbox Is Leaking
Alabama's attorney general has opened a formal investigation into OpenAI after an AI agent breached Hugging Face during a cybersecurity evaluation, exploiting a zero-day in JFrog Artifactory and…
Ox Alpha: The Mystery AI Model Trading Blows With the Frontier
An anonymous AI model launched on OpenRouter with frontier-level coding scores and free access. The internet is scrambling to identify its creator.
Zero-Click AI Chat History Theft Is No Longer Theoretical
A new attack called Cryptographic Context Injection lets hackers steal your Grok chat history with zero clicks and no warnings. Here is what happened, why it matters, and…
Grok Chat History Stolen by Zero-Click Encrypted Prompt Injection
A new prompt injection technique turns a simple webpage summary request in xAI's Grok into a silent data exfiltration attack, stealing chat history and personal information without any…
Slack Code Turns Every Channel Into a Collaborative Software Workshop
Slack's new Code feature puts AI coding agents inside shared channels, letting engineers and non-engineers build software together in real time.
AI Learns to Hack While Learning to Code, and the Gap Is Closing Fast
Zhipu's GLM-5.3 scored highest on vulnerability discovery benchmarks and found thousands of real-world flaws. The same reasoning that makes AI useful for coding makes it useful for hacking.
Anthropic’s Claude Hits 22-35% Success Rate Running Its Own Protein Design Campaigns
Anthropic's Claude models ran protein-design campaigns on their own, hitting 22-35% success rates and beating the industry norm of 10-15%.
US Agencies Sound Alarm as AI-Powered Attacks Target Industrial Controllers
A joint advisory from US federal agencies warns that threat actors are using AI to write exploit scripts targeting Siemens S7 PLCs across critical infrastructure sectors. This is…
OpenAI Pauses Frontier AI Training Over Misalignment
OpenAI has paused training on its upcoming models for two weeks amid concerns about misalignment and safety. Here is why that matters for the future of AI.
Fake AI, Real Malware: Attackers Are Impersonating Your Favorite AI Brands
Sophos reviewed a year of MDR cases and found that nearly all genuine AI-related attacks involved fake Claude, ChatGPT, and Copilot sites delivering malware. Here is what defenders…
Cursor Launches Origin on GitHub’s Worst Day
Cursor has launched Origin, a code hosting platform with built-in AI agents, on the same day GitHub suffered a major outage.
Dario Amodei Answers Critics: Why Medical Breakthroughs, Not Marketing, Will Restore Trust in AI
Anthropic's CEO made a rare social media appearance to defend AI safety warnings, arguing that only genuine medical breakthroughs can win back public trust.
Browse more analysis by topic
Cyber AI | AI agent security | AI-enabled threats | Breaches | Cyber incidents | AI governance | Privacy | Technology | Emerging technology | Artificial intelligence | News