OpenAI Introduces GPT-Red Automated Red Teaming System for AI Safety
OpenAI unveils GPT-Red, an automated red teaming system using self-play to enhance AI safety, alignment, and robustness against prompt injections.
Sources · OpenAI
The news desk
Follow what’s changing.
Recent developments in artificial intelligence, with the original sources one click away.
All news
521–540 of 1,897 Three ArXiv Papers Advance Language Model Capabilities Through Novel Training and Evaluation Methods
New research introduces benchmarks for multimodal judge models, explores behavioral organization in language models, and demonstrates improved exploration techniques.
Sources · arxiv.org
ELIZA: The 1960s Chatbot That Predicted Modern AI Interactions
MIT's 1960s ELIZA chatbot pioneered patterns of human-AI conversation that persist in today's ChatGPT interactions.
Sources · WIRED AI
HUD Withholds Documents on DOGE's AI Use in Housing Policy, Cites Nonexistent Privilege
The Department of Housing and Urban Development has refused to disclose how DOGE used AI for housing policy, citing a privilege that doesn't exist.
Sources · WIRED AI
OpenAI Partners with Kalshi to Display World Cup Betting Odds in ChatGPT
OpenAI integrates Kalshi's prediction market data into ChatGPT to show World Cup odds, marking its first partnership of this kind.
Sources · NYT Technology
OpenAI Publishes Guide on Managing AI Investments in Agentic Era
OpenAI releases guidance for enterprises on managing AI investments by measuring useful work per dollar and scaling high-value workflows.
Sources · OpenAI
OpenAI Reportedly Developing Screenless Speaker with Moving Components
Bloomberg reports OpenAI is working on its first hardware device, a speaker without a screen that features mechanical elements capable of movement.
Sources · TechCrunch AI
Agentic-DPO: New Method Improves AI Agent Training Without Online Environment Interaction
Researchers introduce Agentic-DPO, a lightweight method that converts expert trajectories into state-level preferences, improving agent performance without environment rollouts.
Sources · arxiv.org
Anthropic Publishes New AI Research with Mixed Implications
Anthropic, valued at nearly $1 trillion, releases research with notable findings but unclear broader significance.
Sources · MIT Technology Review
Google and AIM Launch ATL Saathi AI Tool for Indian Robotics Educators
Google and AIM introduced ATL Saathi, a Gemini-powered AI assistant designed to support Indian educators in robotics labs.
Sources · DeepMind
IBM Shares Plunge as AI Spending Shifts from Software to Infrastructure
IBM's stock fell sharply after a Q2 earnings warning, attributing the miss to clients prioritizing AI infrastructure hardware over software.
Sources · Businessinsider / Calcalistech +14
Nested-ReFT Framework Reduces Training Costs for LLM Reasoning Tasks
New method uses off-policy reinforcement learning to cut computational costs in language model fine-tuning while maintaining performance.
Sources · arxiv.org
New Research Addresses Multi-Attribute Steering and Inference Acceleration for Large Language Models
Three recent papers tackle LLM alignment, GPU inference optimization, and multi-agent coordination challenges.
Sources · arxiv.org
New Research Tackles LLM Safety in Critical Infrastructure Through Inference-Time Control
Two ArXiv papers address reliability challenges in LLM systems through structured control mechanisms for safety-critical applications.
Sources · arxiv.org
Researchers Introduce Multi-Agent LLM Systems for Worldbuilding and Quantum Programming
New papers present AutoWorldBuilder for fictional worldbuilding and QAgent for quantum circuit programming, both using multi-agent LLM architectures.
Sources · arxiv.org
Researchers Use Quantum Computing and AI to Develop New Peptides for Rare Diseases
Scientists combined quantum computing with AI to generate peptides for drug development targeting underserved populations and rare diseases.
Sources · WIRED AI
Study Reveals Geopolitical Bias in Major Language Models' Policy Evaluations
Research finds GPT-5, Claude Sonnet, and Gemini rate policies lower when endorsed by China or Russia compared to Western nations.
Sources · arxiv.org
Apple Sues OpenAI Over Alleged Theft of Company Secrets
Apple has filed a lawsuit against OpenAI, accusing the AI company of stealing proprietary information despite their 2024 partnership.
Sources · NYT Technology
OpenAI Targets Family Users with New ChatGPT Product Role
OpenAI is hiring a product manager to develop ChatGPT experiences specifically for families, caregivers, and older adults.
Sources · TechCrunch AI
Amazon SageMaker AI Enables Serverless Fine-Tuning for NVIDIA Nemotron 3 Models
AWS announces serverless customization support for NVIDIA Nemotron 3 models through Amazon SageMaker Studio.
Sources · Amazon AWS AI
