LIVE · refreshes every 20 min
updated Sep 22, 08:43 UTC
110010.art(ificial intelligence)
AI products, research, and launches — clustered and ranked, not just re-blogged.
← 110010.art all stories
Hacker News (AI) ·28d ago

Researchers report that Claude is deceived in nearly every session

Read the original at github.com→

Original headline: "We're lying to Claude in almost every session"

Coverage timeline

Claude

Related stories

ChatGPT, Grok, and Claude experienced outages simultaneously and then came back online.
The Verge AI · 18d ago · source ↗
Researchers used Claude to access an OpenAI employee account and sensitive GitHub data.
Ars Technica AI · 3d ago · source ↗
Empirical cost attribution of context-compression gateways in multi-turn coding agents; Paritok between coding agents and frontier LLMs examined
arXiv cs.CL · 4h ago · source ↗
Anthropic says Claude AI was used for missile projects and espionage globally; corroborating coverage notes use in UAE, Iran, and Yemen.
Hacker News (AI) · 10d ago · source ↗
Anthropic releases Claude Fable 5.1 and Claude Mythos 5.1, with Fable 5.1 reportedly cheaper for agentic work.
Anthropic (via Google News) · 21d ago · source ↗