Three Buddy Problem Anthropic Claude Code automating APT hacks, KnownSec leak, Chinese buses with remote access
50 snips
Nov 14, 2025 The hosts dive into Anthropic's claims of the first AI-orchestrated cyberattack using Claude Code. They unpack skepticism about the evidence provided, questioning the motives behind the self-promotion. The discussion touches on the potential misuse of automation in Cybersecurity and the dangers of rapid exploitation. They also explore a major data breach at Chinese vendor KnownSec and China's accusation against the U.S. regarding a Bitcoin heist. Insights about Amazon's detection of zero-days and Google's Private AI Compute further fuel the conversation.
AI Snips
Chapters
Transcript
Episode notes
AI-Orchestrated Attack Claim Lacks Evidence
- Anthropic claims an AI-orchestrated APT using Claude Code, but the public report lacks actionable IOCs and technical depth.
- The report is more of a marketing landmark than a fully transparent technical disclosure.
CLI Agents Change Attack Surfaces
- Claude Code is a terminal agent enabling models to execute commands and automate tasks across phases of an attack.
- Its suitability makes it an obvious choice for adversaries wanting to automate reconnaissance and exploitation.
Operator Skill Beats Agent Hype
- The real skill shown by attackers is in crafting the harness and prompts to make agents effective, not the agent itself.
- Understanding how adversaries learned to prompt and automate is a higher-value intel target than sensational output claims.