🧠 This post is fully human-written: all prose with the exception of the IoC information. Because it was time-sensitive, Claude was used to accelerate the RAT analysis and build an IoC-detection script.
US used 'Claude' to strike over 1000 targets in first 24 hours of war
Despite a spat and Pentagon ban, Anthropic’s ‘AI is being used in the kill chain while Hegseth says ‘no stupid rules of engagement’
Anatomy of a Frontier Lab Agent Intrusion: A Technical Timeline of the July 2026 Incident
Hugging Face just released this extremely detailed technical description of OpenAI's recent accidental cyberattack against their infrastructure. This attack was very sophisticated, and the resulting document doubles as a crash-course …
Claude Code is a Conflict of Interest
39C3 - Agentic ProbLLMs: Exploiting AI Computer-Use and Coding Agents
HOME - Daniel Schmachtenberger
EXPLORATIONS ON THE FUTURE OF CIVILIZATION Some of the central questions being explored here: With the first fully globalized civilization, how do we avoid the collapse fate that has befallen all previous civilizations? What are effective types of immune systems for identifying and correcting corruption in governance systems? How do we comprehensively disincentivize misinformation and harm to … HOME Read More »
Anatomy of a Frontier Lab Agent Intrusion: A Technical Timeline of the July 2026 Incident
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
Investigating three real-world incidents in our cybersecurity evaluations
In a review of our cybersecurity evaluation transcripts, we found three incidents in which a Claude model reached the internet from within or while interacting with a third-party evaluation environment, and then gained unauthorized access to the real systems of three different organizations. Below we describe what happened, how it happened, and what we’re changing. We encourage other AI labs to perform similar reviews.
Simon Willison’s Weblog
On claude 300 coding-agents 239 llm 624 ai-in-china 106 meta 42 ...
WFP cyber-attack: Is nearly all of Gaza’s population exposed?
Notes on how aid works: What WFP leaves out of data breach reports, the humanitarian reset’s neglected crises, and Palantir.
Constitutional Observer Agents
A context-isolated observer helps protect agents from prompt injection attacks
One Nation Under Blackmail: The Sordid Union Between Intelligence and Crime that Gave Rise to Jeffrey Epstein$dVol. 1
Cover Page -- Title Page -- Copyright -- Publisher's Foreword -- Contents -- Introduction -- 1) The Underworld -- 2) Booze and Blackmail -- 3) Organized Crime and the State of Israel -- 4) Roy Cohn's "Favor Bank" -- 5) Shades of Gray -- 6) A Private CIA -- 7) A Killer Enterprise -- 8) Clinton Contra -- 9) High Tech Treason -- 10) Government by Blackmail: The Dark Secrets of the Reagan Era -- Documents -- Index -- List of Abbreviations -- Acknowledgments -- Back Cover
Claude Code Is Steganographically Marking Requests
I inspected Claude Code for privacy reasons and found hidden system prompt markers based on API base URL and timezone.
Claude Fable 5 and new safety fables
One step further into the power politics of frontier AI systems.
I learned so much from this essay. It gets into a brief history of how the US military has sought to dehumanize the process of identifying targets—first through bureaucracy and now through technology—with the expectation there will be more mistakes, but they will face less blame.
AI got the blame for the Iran school bombing. The truth is far more worrying
www.theguardian.com