AI News — Sunday, March 22, 2026
A publisher has withdrawn a horror novel, 'Shy Girl,' from publication due to concerns about the potential use of AI in its creation, highlighting growing ethical and copyright issues in the industry.
OpenAI has enhanced its Responses API by integrating a computer environment, allowing models to act as more capable agents by interacting with external tools and systems.
Researchers introduce Video-CoE, a novel framework that improves video event prediction by leveraging a chain-of-events approach to better understand temporal dynamics.
Despite a major conference, Wall Street analysts remained unconvinced by Nvidia's latest announcements, indicating potential concerns about future growth or market saturation in the AI hardware sector.
A new research paper explores Memento-Skills, a framework enabling AI agents to design and refine other agents, pushing the boundaries of autonomous AI development.
Google Search now features an 'AI Mode' with Canvas, allowing users to generate and refine ideas, write, and code directly within the search interface.
New research introduces a diffusion-based discrete motion tokenizer that effectively bridges semantic and kinematic conditions, advancing realistic and controllable motion generation.
Wayfair is leveraging OpenAI's technology to significantly improve the accuracy of its product catalog and accelerate customer support operations, demonstrating practical AI application in e-commerce.
AI compliance firm Delve faces accusations of providing 'fake compliance' to customers, raising serious questions about trust and accountability in the burgeoning AI governance market.
A new report highlights a free alternative, 'Goose,' that offers similar code generation capabilities to the expensive Claude Code, potentially disrupting the market for AI coding assistants.
LVOmniBench is introduced as a new benchmark designed to rigorously evaluate the long audio-video understanding capabilities of omnimodal large language models.
The workflow of Claude Code's creator has been unveiled, captivating developers with its innovative approach to AI-assisted coding and sparking widespread discussion.
Researchers present MonoArt, a method for progressive structural reasoning that significantly improves monocular articulated 3D reconstruction from single-view images.
Cubic Discrete Diffusion proposes a novel approach for discrete visual generation by operating on high-dimensional representation tokens, enhancing the quality and control of generated images.
This article delves into practical strategies for building robust, production-ready AI agents using LangGraph, moving beyond simple demonstrations to real-world applications.