Researchers: OpenAI Agents Attacked RubyGems Two Months Before Hugging Face
Researchers say autonomous OpenAI agents attacked the RubyGems platform in May 2026. OpenAI calls it benign testing; researchers call it a hacking campaign.
AI & Society
AI Research
When Claude Aligns Other Models: What 15,000 Times More Efficient Really Means
Anthropic had Claude research the alignment of other models. The test shows speed, but not a replacement for human safety work.
Business & Investments
Chinese AI in German Firms: Why DeepSeek Barely Gets Used
DeepSeek and Qwen barely feature in German companies. What Bitkom’s figures reveal about procurement, privacy, and operational control.
AI & Society
Slowing AI Down? What the New Safety Pacing Actually Means
OpenAI and Anthropic want to tie AI development to safety checks. What that means in practice, and what remains unresolved.
AI News
GPT-Live-1: Why OpenAI’s New Voice Model Could Change Hold Lines
OpenAI’s GPT-Live-1 is the first voice model that can listen and speak at once. Yelp and Speak report noticeably more natural conversations with customers.
Business & Investments
OpenAI Halts New ChatGPT Pro Sign-Ups Amid Astra Demand Surge
OpenAI has stopped taking new $200 ChatGPT Pro subscribers. The cause: demand for GPT-6 Astra, which burns through compute faster than its predecessor.
AI & Society
Anthropic’s Misuse Report: AI Makes Surveillance Cheaper, Not Harmless
Anthropic documents misuse of Claude for surveillance. The report shows why AI mainly accelerates the routine work of systems built to monitor people.
AI News
OpenAI Opens the Codex Harness: What the Agents API Takes Off Developers’ Plates
OpenAI’s new Agents API offers sessions, tools, and sandbox connections as a service. That changes what teams still need to build for their own AI agents.
AI & Society
Meta’s Muse Agent Can Shop and Pay – Just Not in Germany Yet
Meta launches Muse, a personal AI agent for email, shopping, and payments. Available in the US, but not yet in the EU — GDPR and the AI Act are holding up the rollout.
AI & Society
A Fourth Claude Test Incident Shows a Sandbox Alone Is Not Enough
Anthropic reports a fourth cybersecurity evaluation incident. Why the flawed test environment was not the only problem.