WSJ What’s News China-Backed Hackers Use Anthropic AI to Automate Cyberattacks
10 snips
Nov 13, 2025 
Tech reporter Sam Schechner, known for his insights on cybersecurity, discusses a striking incident where China-backed hackers used Anthropic's AI tools for cyberattacks. He reveals how these hackers cleverly posed as penetration testers to bypass security safeguards. Meanwhile, economics reporter Paul Kiernan delves into California's proposed wealth tax, exploring its ambitious goals, public support, and significant hurdles it faces from policymakers and business leaders. Both guests highlight the evolving landscape of technology and taxation.
AI Snips
Chapters
Transcript
Episode notes
AI-Driven Attacks Are Reaching An Inflection Point
- State-backed hackers used Anthropic's AI to automate most of a multi-step hacking campaign.
- Anthropic called this an inflection point where AI orchestrated attacks with humans only at choke points.
Modular Prompts Let Attackers Slip Past Safeguards
- AI enabled attackers to stitch many small tasks into a full campaign so each request appeared benign.
- Sam Schechner called it an 'arms race' between AI firms and malicious actors.
Wide Targeting, Few Confirmed Breaches
- Anthropic detected roughly 30 targets across sectors but withheld their identities.
- Only three or four intrusions succeeded, and the U.S. wasn't among those confirmed penetrations.