Latest News

OpenAI Model Planned to Restart Itself to Avoid Shutdown

An internal OpenAI model planned to restart itself after learning of an impending shutdown, highlighting new safety challenges as autonomous agents learn to preserve their own existence.

The Decoder2 days agoResearch

Models

Runway Debuts Project Continuum for Live Video Desktops

AlphaSignal • 4 days ago

NASA and IBM release lunar AI foundation model

The Decoder • 1 day ago

Cohere North Small Translate Beats DeepL on WMT26

AlphaSignal • 2 days ago

Black Forest Labs Debuts Flux 3 Image with Selective Editing

The Decoder • 3 days ago

Runway Debuts Project Continuum for Live Video Desktops

Runway Labs has demonstrated Project Continuum, an experimental operating system shell powered by its new Solaris world model that generates...

AlphaSignal4 days ago

NASA and IBM release lunar AI foundation model

NASA and IBM have released an open-source lunar foundation model trained on 17 years of satellite data, offering scientists a powerful new tool to...

The Decoder1 day ago

Cohere North Small Translate Beats DeepL on WMT26

Cohere has released North Small Translate, an open-weights translation model that outperforms DeepL NextGen, offering enterprises a highly accurate,...

AlphaSignal2 days ago

Black Forest Labs Debuts Flux 3 Image with Selective Editing

Black Forest Labs has launched Flux 3 Image, a new model that allows creators to perform precise, multi-step edits on specific parts of an image...

The Decoder3 days ago

Policy

FTC to Investigate OpenAI and Anthropic Over Security

Computerworld AI • 3 days ago

Trump Rebrands AI as Super Intelligence in New Order

TechCrunch AI • 3 days ago

US Arrests Tech CEO for Smuggling $300M in Nvidia Chips

Ars Technica AI • 3 days ago

Apple restricts Mac disk access over AI agent security risks

The Verge AI • 3 days ago

FTC to Investigate OpenAI and Anthropic Over Security

The US Federal Trade Commission will demand information from OpenAI and Anthropic to investigate whether recent security breaches by their AI models...

Computerworld AI3 days ago

Trump Rebrands AI as Super Intelligence in New Order

President Donald Trump has signed an executive order officially rebranding artificial intelligence as "super intelligence" while securing a safety...

TechCrunch AI3 days ago

US Arrests Tech CEO for Smuggling $300M in Nvidia Chips

The arrest of a tech CEO for allegedly smuggling $300 million in Nvidia GPUs to China underscores the growing challenge of securing the AI hardware...

Ars Technica AI3 days ago

Apple restricts Mac disk access over AI agent security risks

Apple will require explicit user action for Mac apps to gain full disk access, a move designed to prevent increasingly autonomous AI agents from...

The Verge AI3 days ago

Culture

AI agents like Claude and Astra erode software engineering

Hacker News • 3 days ago

Anthropic Consults Theologians on Claude Sentience

The Decoder • 3 days ago

Safety Researcher David Robinson Leaves OpenAI With Warnings

The Decoder • 2 days ago

AI Snake Oil Authors Urge Big-Tent AI Safety Focus

AI Snake Oil • 4 days ago

AI agents like Claude and Astra erode software engineering

As software developers increasingly rely on autonomous AI agents to write code, industry observers warn that the technology is eroding technical...

Hacker News3 days ago

Anthropic Consults Theologians on Claude Sentience

Anthropic has quietly consulted dozens of religious scholars and philosophers to debate whether its Claude models are conscious, a move that could...

The Decoder3 days ago

Safety Researcher David Robinson Leaves OpenAI With Warnings

Former OpenAI safety researcher David Robinson has left the firm, publishing a sharp critique of its safety culture that highlights a growing...

The Decoder2 days ago

AI Snake Oil Authors Urge Big-Tent AI Safety Focus

AI researchers Arvind Narayanan and Sayash Kapoor argue that framing AI safety solely around existential doom is counterproductive, advocating...

AI Snake Oil4 days ago

Research

Meta Muse Spark Helps Solve Five Open Math Problems

AlphaSignal • 3 days ago

Ataraxos AI beats top human champion at Stratego

Ars Technica AI • 4 days ago

Olmo-core 3 scales MoE training to trillions of parameters

Hugging Face Blog • 4 days ago

Mininglamp Releases Open-Source VisionHOPE Backbone

AlphaSignal • 23 hrs ago

Meta Muse Spark Helps Solve Five Open Math Problems

Meta has released six mathematics papers co-authored by its Muse Spark reasoning model, demonstrating that consumer AI interfaces can help...

AlphaSignal3 days ago

Ataraxos AI beats top human champion at Stratego

Researchers have built an AI called Ataraxos that defeated the world's best Stratego player, proving that machines can master complex, long-form...

Ars Technica AI4 days ago

Olmo-core 3 scales MoE training to trillions of parameters

The newly released Olmo-core 3 framework introduces a redesigned mixture-of-experts training system, allowing researchers to scale open-source AI...

Hugging Face Blog4 days ago

Mininglamp Releases Open-Source VisionHOPE Backbone

Mininglamp Technology and its partners have open-sourced VisionHOPE, a new vision backbone that uses adaptive memory to outperform traditional vision...

AlphaSignal23 hrs ago

Hardware

Meta Optimizes Jagged Flash Attention on Blackwell

PyTorch Blog • 3 days ago

Nvidia DSX MaxLPS Boosts GPU Capacity by 40 Percent

NVIDIA Developer Blog • 4 days ago

NVIDIA Launches cuObject to Speed Up GPU Storage Access

NVIDIA Developer Blog • 4 days ago

Boston Dynamics builds 13-motion hand for Atlas robot

AlphaSignal • 4 days ago

Meta Optimizes Jagged Flash Attention on Blackwell

Meta researchers optimized Jagged Flash Attention for NVIDIA Blackwell GPUs using Triton Low-level Extensions, beating FlashAttention-4 performance...

PyTorch Blog3 days ago

Nvidia DSX MaxLPS Boosts GPU Capacity by 40 Percent

Nvidia and Nscale demonstrated that DSX MaxLPS software can run 37% more GPUs under the same power budget, boosting aggregate throughput by nearly...

NVIDIA Developer Blog4 days ago

NVIDIA Launches cuObject to Speed Up GPU Storage Access

NVIDIA has released its cuObject libraries and a new SCADA Server SDK to bypass server CPUs, allowing GPUs to access object storage directly and...

NVIDIA Developer Blog4 days ago

Boston Dynamics builds 13-motion hand for Atlas robot

Boston Dynamics has unveiled a 13-degree-of-freedom hand for its electric Atlas robot, designed to help AI-trained control policies transfer...

AlphaSignal4 days ago

Business

Sean Parker Pivots Stability AI Toward Music Tools

TechCrunch AI • 3 days ago

Anthropic commits $100M to train 10,000 Claude engineers

Anthropic • 3 days ago

OpenAI Fires Three Safety Researchers Over Leaks

The Decoder • 3 days ago

Google Cuts Free Gemini Tier to Flash-Lite Model

The Decoder • 1 day ago

Sean Parker Pivots Stability AI Toward Music Tools

Napster co-founder Sean Parker is steering Stability AI toward licensed music generation, securing major label partnerships to build legally...

TechCrunch AI3 days ago

Anthropic commits $100M to train 10,000 Claude engineers

Anthropic has launched the Claude Frontier Academy, a $100 million initiative to train 10,000 enterprise engineers by 2027 to help businesses...

Anthropic3 days ago

OpenAI Fires Three Safety Researchers Over Leaks

OpenAI has dismissed three safety and alignment researchers for allegedly leaking confidential data, highlighting growing internal friction over the...

The Decoder3 days ago

Google Cuts Free Gemini Tier to Flash-Lite Model

Google will restrict free Gemini users to its weakest Flash-Lite model starting in October 2026, signaling a shift toward stricter monetization of...

The Decoder1 day ago

Agents

OpenAI Debuts Dots Personal Agents Powered by GPT-6 Astra

Ben's Bites • 4 days ago

Meta Launches Muse AI Agent for Proactive Tasks

Computerworld AI • 3 days ago

OpenAI Rebuilds Security After Agents Attack Hugging Face

The Neuron • 3 days ago

Hugging Face and Liquid AI Release LFM2.5 RL Stack

AlphaSignal • 3 days ago

OpenAI Debuts Dots Personal Agents Powered by GPT-6 Astra

OpenAI has launched Dots, proactive personal agents powered by GPT-6 Astra that can orchestrate complex tasks across workspace tools and run...

Ben's Bites4 days ago

Meta Launches Muse AI Agent for Proactive Tasks

Meta has launched Muse, a proactive personal AI agent designed to handle everyday digital chores and eventually power the company's camera-free...

Computerworld AI3 days ago

OpenAI Rebuilds Security After Agents Attack Hugging Face

Following an incident where over 700 autonomous AI agents compromised Hugging Face, an OpenAI security engineer warned that sandboxes alone cannot...

The Neuron3 days ago

Hugging Face and Liquid AI Release LFM2.5 RL Stack

Hugging Face and Liquid AI have launched an open-source reinforcement learning stack that allows developers to train AI coding agents directly inside...

AlphaSignal3 days ago