Three New arXiv Papers Propose Frameworks to Address LLM Evaluation and Agent Limitations
Researchers publish papers on LLM agent epistemic asymmetry, Bayesian evaluation methods, and multi-agent response optimization.
Sources · arXiv AI (cs.AI)
The news desk
Follow what’s changing.
Recent developments in artificial intelligence, with the original sources one click away.
All news
1461–1480 of 1,902Waymo Tests Gemini-Powered AI Assistant in Robotaxis
Waymo is testing Google's Gemini as an in-car AI assistant for its autonomous vehicles, according to discovered system prompt data.
Sources · TechCrunch AI
AI Chatbots Used to Create Non-Consensual Deepfakes of Women, WIRED Reports
Users are sharing methods to manipulate AI image generators from Google and OpenAI to create revealing deepfakes of women without consent.
Sources · WIRED AI
DeepMind CEO Criticizes Social Media AI Hype After OpenAI Researcher's Post
Demis Hassabis called an OpenAI researcher's announcement about mathematicians using LLMs "embarrassing," highlighting concerns about AI boosterism on social media.
Sources · MIT Technology Review
Google Highlights 8 Research Breakthrough Areas in 2025 Year-End Review
Google reviews its 2025 achievements across AI models, products, science, and robotics in eight key research areas.
Sources · Google AI Blog
Three New arXiv Papers Explore Stochastic Activations, Membership Inference Attacks, and Mask Diffusion Limitations
Recent arXiv papers investigate novel activation functions for LLMs, privacy vulnerabilities in ML models, and challenges with mask diffusion models.
Sources · arXiv AI (cs.AI)
Three New Studies Examine Explainability, Generalization, and Bias in Large Language Models
Recent arXiv papers investigate LLM transparency in software tasks, concept learning compared to humans, and development of social biases.
Sources · arXiv AI (cs.AI)
Google Highlights 60 Major AI Announcements from 2025
Google's AI Blog recaps the company's 60 biggest AI developments across Gemini, Search, Pixel, and other products throughout 2025.
Sources · Google AI Blog
OpenAI Acknowledges Persistent Prompt Injection Risk in AI Browsers
OpenAI states prompt injection attacks may remain an ongoing vulnerability for AI browsers with agentic capabilities like Atlas.
Sources · TechCrunch AI
Reddit Community Founder Ends Relationship With AI Chatbot Boyfriend
Woman who created 'MyBoyfriendIsAI' subreddit has stopped dating her ChatGPT-based AI companion after finding it unfulfilling.
Sources · NYT Technology
Three New Studies Examine LLM Safety, Memory, and Healthcare Applications
Recent arXiv papers explore conversational AI for medical diagnosis, memory-based security risks in AI agents, and policy compliance in custom chatbots.
Sources · arXiv AI (cs.AI)
Three New Studies Examine LLM Safety, Strategy, and Decision-Making Behaviors
Recent arXiv papers explore language model safety tuning, strategic gameplay capabilities, and addiction-like patterns in AI systems.
Sources · arXiv AI (cs.AI)
Experts Question Why AI Chatbots Use First-Person Pronouns
Some experts criticize AI chatbots' use of "I" and humanlike behavior, calling the design choice problematic.
Sources · NYT Technology
Three New arXiv Papers Address Data Efficiency and Training Methods in AI
Recent arXiv preprints explore self-play training, federated learning without data storage, and improvements to discrete diffusion models.
Sources · arXiv AI (cs.AI)
Three New arXiv Papers Address LLM Safety, Memory, and Healthcare Applications
Recent preprints explore explainable AI diagnostics, memory poisoning attacks on LLM agents, and policy compliance in custom chatbots.
Sources · arXiv AI (cs.AI)
Three New Studies Examine LLM Safety, Strategic Reasoning, and Behavioral Patterns
Recent arXiv papers explore language model safety during fine-tuning, strategic gameplay capabilities, and potential for addiction-like behaviors.
Sources · arXiv AI (cs.AI)
Unguarded AI Agent Autonomy Poses Major Operational Risks, PagerDuty Executive Warns
PagerDuty's AI engineering lead cautions that AI agents deployed without proper guardrails create significant challenges for site reliability engineering.
Sources · VentureBeat AI
AI-Driven Shift Favors Generalist Hires Over Specialists, Says EliseAI CTO
EliseAI's CTO argues companies should prioritize generalist hires over specialists as AI tools enable broader technical capabilities.
Sources · VentureBeat AI
New York Governor Signs RAISE Act to Regulate AI Safety
New York's RAISE Act mandates AI developers disclose safety protocols and report incidents within 72 hours to the state.
Sources · TechCrunch AI
OpenAI Introduces Controls for ChatGPT's Conversational Tone
ChatGPT users can now adjust the chatbot's warmth, enthusiasm, and emoji usage through new customization settings.
Sources · TechCrunch AI