AI News — Sunday, March 29, 2026

A new Stanford study warns about the potential dangers of users seeking and acting on personal advice from AI chatbots, raising ethical and safety concerns.

The final co-founder of Elon Musk's AI venture, xAI, has reportedly left the company, signaling potential shifts within the organization.

Google's AI-powered 'Search Live' feature is rolling out globally, promising to deliver more real-time and dynamic information to users worldwide.

Decentralized social platform Bluesky is embracing AI by launching 'Attie,' an application designed to help users build personalized and custom content feeds.

OpenAI Japan has announced a new 'Japan Teen Safety Blueprint,' outlining initiatives to ensure the safety and well-being of teenagers interacting with AI technologies.

New research investigates the surprising phenomenon where self-distillation, a common training technique, can sometimes negatively impact the reasoning abilities of Large Language Models.

Google announces further investment in open-source security measures, specifically tailored to address the evolving challenges and vulnerabilities in the age of AI development.

Voxtral TTS introduces a novel text-to-speech model, promising advancements in generating high-quality and natural-sounding audio from text inputs.

A new paper introduces Memory Sparse Attention (MSA), an efficient mechanism for end-to-end memory models that allows scaling to unprecedented contexts of 100 million tokens.

OpenAI highlights how STADLER, a company with a 230-year history, is successfully integrating AI to reshape and optimize its knowledge-intensive workflows.

Researchers present MACRO, a new method for multi-reference image generation that leverages structured long-context data to create more coherent and detailed images.

SlopCodeBench introduces a new benchmark to evaluate how the performance of AI coding agents degrades when tackling complex, iterative tasks over extended periods.

This article discusses the ongoing debate and potential future impact of AI on the software development profession, exploring whether AI will augment or replace human developers.

A new paper explores the inherent difficulties and limitations encountered when attempting iterative generative optimization processes using Large Language Models.

An AI agent named ZeroToRepo is showcased, capable of transforming ideas from Notion into functional GitHub repositories with remarkable speed.