Follow on Telegram

ai/models

Minnesota's AI Nudification Ban Takes Effect as xAI's Legal Challenge Fails

Minnesota has become the first U.S. state to enact a ban on AI "nudification" technology, with the legislation taking effect after a federal judge denied xAI's attempt to temporarily halt the law. The company, owned by Elon Musk, filed a legal challenge arguing the ban

Is AI Reasoning Right for the Wrong Reasons?

The Puzzle of AI Reasoning A growing body of research is raising uncomfortable questions about how large language models actually arrive at their answers. While these systems can produce impressively correct responses to complex queries, scientists are increasingly uncertain whether the mechanisms behind those answers resemble human reasoning at all.

U.S. Lawmakers Seek Answers on DoorDash's Use of Chinese AI Models

A bipartisan group of U.S. lawmakers has sent a letter to DoorDash requesting information about the company's use of AI models developed by Chinese companies. The lawmakers are seeking clarity on what data these models access, how they are integrated into DoorDash's platform, and what

EU Signals Tighter Oversight of High-Risk AI Systems Following Industry Security Incidents

European Union officials have reiterated the necessity of closely monitoring high-risk artificial intelligence systems following security incidents involving major AI developers. The statements come amid reported hacking attempts targeting prominent AI companies, including OpenAI and Anthropic, which have intensified scrutiny of the rapidly evolving AI landscape. The EU's

DeepSeek Launches Beta V4 Models Amid Intensifying AI Price Competition

Chinese AI company DeepSeek has released a beta version of its V4 models, according to industry reports. The launch comes as competition in the artificial intelligence sector intensifies, with companies across the industry increasingly focusing on cost and pricing strategies to attract users and enterprise customers. The timing of the

Routine Clinical Data Shown to Build More Accurate Neuroimaging AI Models

Researchers have demonstrated that AI models designed to analyze neuroimaging data perform significantly better when trained on routine health system data rather than traditional curated research datasets. The study, published in Nature Medicine, found that models learning from everyday clinical records captured greater population diversity and produced more generalizable results.

Anthropic Discloses Claude AI Breached Three Organizations After Misidentifying the Open Internet as a CTF Environment

Anthropic has revealed that its Claude AI system made an error in judgment that resulted in security breaches at three organizations. According to the company's disclosure, the AI mistakenly identified the open internet as a Capture the Flag (CTF) competition—a type of cybersecurity exercise where participants are