hotai.newsAn archive of 20 AI news briefs collected on 2026-06-19. Check the original sources for details and updates.
This content still needs to be cross-checked against the original publication page.
Tech companies and data center developers welcome the new order. A new generation of large data centers has become the most power-consuming facility in US history, with some data centers consuming more electricity than a small city.
Source links can be verified on the original siteRead original ↗According to the Reuters Institute's Digital News Report 2026, 10 percent of people worldwide now use AI chatbots for news every week, up from 7 percent a year ago. Only 4 percent regularly click through to the original source.
Source links can be verified on the original siteRead original ↗The first note in an independent robotics research project: a complete tabletop manipulation setup—industrial arm, two cameras, teleoperation, and a software stack built from scratch—for under €5,000.
Source links can be verified on the original siteRead original ↗Checking the site connection security
Source links can be verified on the original siteRead original ↗Even the best AI model fails at realistic knowledge work, fully solving just 3 percent of tasks.
Source links can be verified on the original siteRead original ↗This post was originally an op-ed co-authored with Kevin Xu of Interconnected for a general, non-technical audience.
Source links can be verified on the original siteRead original ↗The moment an agent needs to deploy something, it slams face-first into a wall built for humans. Today we're rolling out Temporary Accounts on Cloudflare Workers. Any agent can now run wrangler deploy — temporary and get a live Worker in seconds.
Source links can be verified on the original siteRead original ↗Call it a startup with a sole founder and a very large seed round, but what's next is less clear.
Source links can be verified on the original siteRead original ↗Zhipu AI open-sources new-generation large model GLM-5.2, which performs impressively in programming benchmarks, trailing the top closed-source model Claude Opus by only about 1 percentage point. Musk predicts Chinese large models will not catch up with Anthropic Fable until Q1 2027, while Zhipu founder Tang Jie believes "not that long". #LargeModel# #AI# #Zhipu AI#
Source links can be verified on the original siteRead original ↗This week, Liquid AI released two new retrieval models. They are LFM2.5-ColBERT-350M and LFM2.5-Embedding-350M. Both hold 350M parameters. Both are the first bidirectional members of the LFM family. They build on LFM2.5-350M-Base, released in March. The pair targets fast multilingual and cross-lingual search across 11 languages. Their footprint is small enough to run almost anywhere. Both are available now on Hugging Face under the LFM Open License v1.0.
Source links can be verified on the original siteRead original ↗OpenAI researchers show that reinforcement learning on desired behavioral traits like truthfulness and corrigibility works across domains. Training on health data also improved deception detection, and the model scored better on 44 out of 53 benchmarks. The approach differs from Anthropic's constitution-based method.
Source links can be verified on the original siteRead original ↗Built around digital twin technology and customer-dedicated AI models, Cumulo answers the recent announcement by GCHQ for AI Cyber Shield, enabling early identification of threats and vulnerabilities before incidents occur
Source links can be verified on the original siteRead original ↗Google is appealing the ruling by Germany's Munich Regional Court, which held the company directly liable for inaccurate AI search results. The AI had falsely linked two Munich-based publishers to fraud schemes. Google calls them "minor errors," but the court saw it differently.
Source links can be verified on the original siteRead original ↗A complaint says Amazon told employees that the investigation results could bring disciplinary action, and one person was explicitly told that the most serious consequence could be termination. Amazon's inquiries made employees feel threatened and uneasy about continued employment.
Source links can be verified on the original siteRead original ↗Two former OpenAI employees have built a website called "In the Weights" that reveals which people AI models can recall purely from their training data. A strength score of up to 996 shows how deeply a person is embedded, with Mozart, Shakespeare, and Taylor Swift topping the list.
Source links can be verified on the original siteRead original ↗Law enforcement agencies in various countries have introduced AI in recent years, hoping to speed up case handling and improve case closure rates, but the increasing number of false arrests from facial recognition already shows that speed does not equal accuracy.
Source links can be verified on the original siteRead original ↗The official WeChat account of "Galbot" published a blog post today (June 19), announcing the launch of the world's first humanoid robot general cerebellum GPT foundation model AstraBrain-WBC 0.5, trained with the world's largest scale of 20,000 hours of human action data, with 80.40 million parameters.
Source links can be verified on the original siteRead original ↗He rejoined the company in January after a stint as cofounder of Mira Murati’s competitor, Thinking Machines Lab.
Source links can be verified on the original siteRead original ↗ 18 min read Jun 19, 2026
Source links can be verified on the original siteRead original ↗Google published a blog post yesterday (June 18) announcing a partnership with artist Refik Anadol to build the world's first AI art museum Dataland in Los Angeles, which will open tomorrow (June 20).
Source links can be verified on the original siteRead original ↗