AI News — Sunday, May 31, 2026
SoftBank announced a massive €75 billion investment to construct new data centers in France, signaling a major commitment to AI infrastructure in Europe.
GitHub Copilot's shift to a token-based billing model has caused significant frustration and consternation among developers due to unexpected cost increases.
Meta is reportedly working on an AI-powered pendant, indicating a potential new direction for wearable AI devices and human-computer interaction.
Nous Research has released NousCoder-14B, an open-source coding model, positioning it as a free and competitive alternative to expensive commercial tools like Claude Code.
New research delves into how vision-language models represent spatial relationships, offering insights into their understanding of object positions and interactions.
Listen Labs secured $69 million in funding to expand its AI-driven customer interview platform, following a successful viral billboard hiring campaign.
Researchers introduce UniSteer, a novel method using text-guided flow matching in activation space to enable more versatile and precise steering of large language models.
As the Hermes AI agent becomes more intelligent and capable, developers are noting a significant increase in its operational costs, raising concerns about long-term affordability.
Google's AI Blog recaps 12 significant announcements and developments from the I/O 2026 keynote, emphasizing advancements in AI and technology.
A new paper proposes LoMo, a technique for local modality substitution that aims to achieve deeper and more effective fusion between vision and language models.
OpenAI has published a shared playbook outlining best practices for conducting trustworthy third-party evaluations of AI systems, focusing on safety and reliability.
An article argues that AI agents should be proactive in initiating communication with users, rather than solely reacting to prompts, to enhance user experience and utility.
New research introduces Skill0.5, a method for agentic reinforcement learning that combines skill internalization and utilization to improve generalization in out-of-distribution scenarios.
A paper explores the challenges and insights gained from integrating cloud-based AI agents with device-side agents in hybrid multi-agent systems.
An article highlights 'inference theft' as a critical new security vulnerability for AI applications and provides guidance on how to protect large language model endpoints.