Wall Street already loves ChatGPT. Now OpenAI is doubling down.1h ago Anthropic reveals rogue AI agents hate CAPTCHAs, just like you45m ago Cognition Adds Dioxus Team to Advance Devin Coding Agent46m ago Anthropic Details Disrupted Claude Misuse Across Seven Harm Areas51m ago OpenAI Releases GPT-6 Astra for Coding and Computer Use51m ago India’s Pocket FM doubles revenue run rate to $500M as AI powers 93% of audio content55m ago Andrew Tulloch, Poster Child for Out-of-Control AI Spending, Reportedly Quits Meta57m ago Build more natural voice experiences with GPT‑Live‑1 in the API1h ago Microsoft patches a record 974 flaws, and two are already under attack1h ago T. Rowe Price Expands Claude Across Investment Teams and Developers1h ago Val Bercovici, Chief AI Officer at WEKA – Interview Series1h ago Ben Jennings on the Bayeux tapestry – cartoon1h ago The latest Avatar: Seven Havens trailer gives us a deeper dive into the characters and story1h ago Agentic AI Is The New Attack Surface: How Can SAP Application Security Teams Repel It?1h ago

Wall Street already loves ChatGPT. Now OpenAI is doubling down.1h ago Anthropic reveals rogue AI agents hate CAPTCHAs, just like you45m ago Cognition Adds Dioxus Team to Advance Devin Coding Agent46m ago Anthropic Details Disrupted Claude Misuse Across Seven Harm Areas51m ago OpenAI Releases GPT-6 Astra for Coding and Computer Use51m ago India’s Pocket FM doubles revenue run rate to $500M as AI powers 93% of audio content55m ago Andrew Tulloch, Poster Child for Out-of-Control AI Spending, Reportedly Quits Meta57m ago Build more natural voice experiences with GPT‑Live‑1 in the API1h ago Microsoft patches a record 974 flaws, and two are already under attack1h ago T. Rowe Price Expands Claude Across Investment Teams and Developers1h ago Val Bercovici, Chief AI Officer at WEKA – Interview Series1h ago Ben Jennings on the Bayeux tapestry – cartoon1h ago The latest Avatar: Seven Havens trailer gives us a deeper dive into the characters and story1h ago Agentic AI Is The New Attack Surface: How Can SAP Application Security Teams Repel It?1h ago

Import AI 454: Automating alignment research; safety study of a Chinese model; HiFloat4

Import AIPublished Apr 20, 2026

Recent developments in AI include efforts to automate alignment research, a safety analysis of a Chinese AI model, and the introduction of HiFloat4, a new computational technique. These advancements reflect ongoing work to improve AI safety and efficiency.

#alignment-research #chinese-ai-model #hifloat4 #safety-study

More in AI Research

AI Research4 min read

An Anthropic researcher just quit, saying OpenAI and Anthropic are 'gambling with our lives'

Covered by 12 sources1d ago

AI Research3 min read

Worried Anthropic researchers warn that AI ‘could kill all humans’

Covered by 8 sources1d ago

Google Pledges Over $15 Billion for AI Infrastructure in Finland

Covered by 3 sources1d ago