California Gubernatorial Candidate Proposes Jobs Guarantee for AI-Displaced Workers
Tom Steyer proposes new jobs guarantee program to protect California workers displaced by artificial intelligence.
Sources · WIRED AI
The news desk
Follow what’s changing.
Recent developments in artificial intelligence, with the original sources one click away.
All news
841–860 of 1,902OpenAI Details Security Measures for Running Codex Coding Agent
OpenAI outlines its security infrastructure for Codex, including sandboxing, approval workflows, and telemetry to enable safe AI coding agent deployment.
Sources · OpenAI
OpenAI Responds as Musk-Altman Trial Enters Second Week
OpenAI fired back in court as Elon Musk's lawsuit against the company continued, with new details emerging about Musk's past recruitment efforts.
Sources · MIT Technology Review
Researchers Address Critical Gaps in Large Language Model Evaluation and Safety
New studies tackle LLM robustness, clinical applications, quantum enhancement, and spatial reasoning benchmarks.
Sources · arxiv.org
Researchers Propose Multi-Agent AI Frameworks for Specialized Reasoning Tasks
New research explores multi-agent AI systems for single-cell annotation, geo-spatial reasoning, and retrieval-augmented generation.
Sources · arxiv.org
Chrome Users Discover 4-GB Google AI Model Installation, Prompting Privacy Concerns
Google quietly integrated a 4-GB Gemini AI model into Chrome, surprising users and raising privacy questions, though removal options exist.
Sources · WIRED AI
New Research Reveals Social Alignment Gaps in Large Language Model Reward Systems
Study finds reward models used to align LLMs often prefer socially undesirable responses, revealing critical gaps in current evaluation methods.
Sources · arxiv.org
OpenAI Introduces Three New Voice Models to Its API
OpenAI launches GPT-Realtime-2, GPT-Realtime-Translate, and GPT-Realtime-Whisper to enable developers to build more natural voice applications.
Sources · openai.com / techcrunch.com
OpenAI Introduces 'Trusted Contact' Feature to Address Self-Harm Concerns
OpenAI launches new safety feature allowing ChatGPT users to designate a trusted contact for self-harm situations.
Sources · TechCrunch AI
Research Finds LLM 'Thinking Mode' Has Minimal Impact on Moral Judgments Despite Reducing Cross-Model Disagreement
Study of five frontier models shows reasoning mode doesn't significantly change moral verdicts but reduces cross-model disagreement in disputed scenarios.
Sources · arxiv.org
DeepSeek Eyes $45B Valuation in First Investment Round
Chinese AI lab DeepSeek could reach $45 billion valuation in its inaugural funding round after gaining attention for cost-efficient models.
Sources · TechCrunch AI
Musk Attempted to Recruit Altman to Tesla AI Lab Before Leaving OpenAI Board
Court documents reveal Musk offered Altman a Tesla board seat and AI lab position months before departing OpenAI in 2018.
Sources · wired.com / techcrunch.com
New Research Reveals Social Alignment Failures in LLM Reward Models
Study finds reward models used in AI alignment often prefer socially undesirable responses across bias, safety, and ethical domains.
Sources · arxiv.org
Study Finds LLM 'Thinking Mode' Changes Ethical Frameworks More Than Judgments
New research shows reasoning mode affects how AI models justify moral decisions but rarely changes the actual verdicts across five frontier LLMs.
Sources · arxiv.org
Tomofun Deploys Vision-Language Models on AWS Inferentia2 for Pet Behavior Detection
Pet-tech startup Tomofun uses AWS Inferentia2 chips to reduce costs while maintaining accuracy in its Furbo Pet Camera's AI detection capabilities.
Sources · Amazon AWS AI
AWS Introduces Capacity-Aware Instance Fallback for SageMaker AI Endpoints
Amazon SageMaker AI now automatically switches between instance types when GPU capacity is unavailable, eliminating manual endpoint configuration retries.
Sources · aws.amazon.com / arxiv.org
Elon Musk's Lawyers Question OpenAI President's $30 Billion Valuation at Federal Trial
Musk's legal team challenged Greg Brockman's motivations during OpenAI trial, while texts revealed settlement attempts.
Sources · nytimes.com / techcrunch.com
ML-Agent: New Framework Trains 7B Model to Match GPT-5 Performance on Autonomous Machine Learning Tasks
Researchers develop reinforcement learning framework that trains smaller language models for autonomous ML engineering at lower computational cost.
Sources · arxiv.org
Musk v. Altman Trial Begins in Oakland as OpenAI Lawsuit Gets Underway
Elon Musk and Sam Altman faced off in court last week as Musk's lawsuit against OpenAI commenced in Oakland, California.
Sources · MIT Technology Review
Researchers Advance LLM Agent Memory Systems with Episodic and Semantic Approaches
New research papers introduce memory architectures enabling LLM agents to learn from examples and maintain context over extended interactions.
Sources · arxiv.org