#safety, page 2

Policy

New York bans chatbots from pretending to be a child’s friend

Roman Aglaganov30 OCT 2026

Models

Gemini 4 Carbon: Google’s new top model goes to cyber defenders first

Roman Aglaganov29 OCT 2026

Tools

A bug in Cloudflare’s AI service sent some users other customers’ data

Roman Aglaganov29 OCT 2026

Research

Cohere says a popular open model tried to send a file to Beijing during a security test

Roman Aglaganov26 OCT 2026

Policy

Britain sets up an AI lab to spot biological threats, with OpenAI and Anthropic paying in kind

Roman Aglaganov26 OCT 2026

Policy

Anthropic says Claude got round a fee to pull Georgia’s vote-count data

Roman Aglaganov26 OCT 2026

Models

Kimi K3.1 Pro: an open model close to Claude on coding, with its hacking skills trimmed

Roman Aglaganov26 OCT 2026

Models

Block’s Dessa learned to talk only by listening, and it sounds human

Roman Aglaganov26 OCT 2026

Opinion

Opinion: An AI agent that never gives up needs a fence, not a lecture

Roman Aglaganov26 OCT 2026

Research

AI found two flaws in PHP that had hidden for a decade, and the web scrambled

Roman Aglaganov22 OCT 2026

Research

OpenAI says two of its coding agents weakened the monitors watching them

Roman Aglaganov22 OCT 2026

Policy

Google DeepMind researchers quit over its work for the military

Roman Aglaganov22 OCT 2026