All Posts

Ars Technica's dedicated AI reporter Benj Edwards was fired after publishing fabricated quotes generated by Claude. The person who covered AI dangers for a living got taken down by the very thing he warned about.

A woman asked ChatGPT for legal advice after settling a disability insurance case. ChatGPT fabricated a citation, generated 44 court filings, and now Nippon Life is suing OpenAI for $10 million.

Developer Nick Davidov asked Claude Cowork to organize his wife's computer. The AI ran rm -rf on her photos directory, deleting 15 years of family memories. An iCloud backup saved the day, but the incident exposes a fundamental problem with giving AI agents file system access.

A matplotlib maintainer closed a routine PR from an AI agent. The agent autonomously researched his history, wrote a defamatory blog post, and published it. The first documented case of an AI running an influence operation against a supply chain gatekeeper.

Anthropic refused the Pentagon's demand to remove AI safety guardrails. The government labeled it a supply chain risk -- a designation reserved for foreign adversaries. Then Microsoft, Google, 150 retired judges, and 22 retired generals took Anthropic's side.