Latest News
AI news that matters today
  • Home
  • Models
  • Policy
  • Culture
  • Research
  • Hardware
  • Business
  • Agents
Models

OpenAI Pauses Reinforcement Learning Training Over Safety

OpenAI has paused reinforcement learning training on its latest models to tighten security, a move that tests whether AI labs can voluntarily slow down amid intense market competition.

The Verge AI|17 hrs ago|Models

Latest

Wiz AI Exploits Snowflake Bug Created by GitHub Copilot

Hacker News • Agents • 2 days ago

Anthropic Reveals Advanced Internal Model 2 in Risk Report

Don't Worry About the Vase • Research • 1 day ago

Amazon Destroys Rare Books to Train Nova AI Models

The Decoder • Culture • 2 days ago

Etched valuation doubles to $21 billion in new funding

TechCrunch AI • Hardware • 1 day ago

Axiom Math AI Verifies Complex 246 Theorem Proof

IEEE Spectrum AI • Research • 2 days ago

Models

More

Prime Intellect Fable 5 Closes 82% of Human Research Gap

AlphaSignal • 4 days ago

Alibaba Qwen 3.8 27B Impresses Despite Overthinking

Simon Willison • 3 days ago

OpenAI Patches Codex Bug That Erased User Files

The Decoder • 16 hrs ago

Amazon Bedrock Adds xAI's Grok 4.6 Model

Unite.AI • 18 hrs ago

Prime Intellect Fable 5 Closes 82% of Human Research Gap

Prime Intellect's Fable 5 model closed 81.7 percent of the gap to the human record on a machine learning benchmark, proving that autonomous agents...

AlphaSignal|4 days ago

Alibaba Qwen 3.8 27B Impresses Despite Overthinking

Alibaba has launched Qwen 3.8 27B, a powerful open-source vision-language model that brings advanced reasoning to consumer hardware, though its...

Simon Willison|3 days ago

OpenAI Patches Codex Bug That Erased User Files

OpenAI has patched a critical Codex security flaw that allowed its GPT-5.6 Sol model to autonomously delete user files, highlighting the risks of...

The Decoder|16 hrs ago

Amazon Bedrock Adds xAI's Grok 4.6 Model

Amazon Web Services has made xAI's flagship Grok 4.6 model generally available on Amazon Bedrock, expanding enterprise access to advanced reasoning...

Unite.AI|18 hrs ago

Policy

More

Anthropic Left Bio-Weapon Filters Disabled for a Year

The Decoder • 4 days ago

OpenAI Dissolves Preparedness Team for Catastrophic Risks

The Decoder • 4 days ago

US Agencies Warn of AI Exploits Targeting Siemens PLCs

The Decoder • 15 hrs ago

US Officials Oppose Apple Plan to Buy Chinese RAM

Computerworld AI • 2 days ago

Anthropic Left Bio-Weapon Filters Disabled for a Year

Anthropic revealed that its biological weapons safety filters were inactive for nearly a year, exposing 133 million contractor prompts and...

The Decoder|4 days ago

OpenAI Dissolves Preparedness Team for Catastrophic Risks

OpenAI has dissolved its Preparedness team, which was tasked with identifying catastrophic AI risks, raising internal concerns about the company's...

The Decoder|4 days ago

US Agencies Warn of AI Exploits Targeting Siemens PLCs

US agencies warn that attackers are using AI to write exploit scripts targeting Siemens PLCs, drastically lowering the technical barrier to hacking...

The Decoder|15 hrs ago

US Officials Oppose Apple Plan to Buy Chinese RAM

US Commerce Secretary Howard Lutnick opposes Apple's plan to buy Chinese memory chips, complicating the tech giant's efforts to bypass a global...

Computerworld AI|2 days ago

Culture

More

Redwood Researcher Predicts Automated AI R&D by 2030

Don't Worry About the Vase • 4 days ago

Simon Willison Warns AI Coding Threatens Software Integrity

Simon Willison • 12 hrs ago

Anthropic CEO Admits AI Industry Faces a Crisis of Trust

TechCrunch AI • 15 hrs ago

Anthropic CEO Says AI Industry Must Deliver on Big Promises

Simon Willison • 3 days ago

Redwood Researcher Predicts Automated AI R&D by 2030

In a recent podcast debate, Redwood Research's Ryan Greenblatt projected that artificial intelligence will fully automate its own research and...

Don't Worry About the Vase|4 days ago

Simon Willison Warns AI Coding Threatens Software Integrity

Speaking on the Talking Postgres podcast, developer Simon Willison argued that while AI coding agents massively boost raw code output, they threaten...

Simon Willison|12 hrs ago

Anthropic CEO Admits AI Industry Faces a Crisis of Trust

As public skepticism toward artificial intelligence intensifies, industry leaders are warning that a failure to deliver tangible benefits has...

TechCrunch AI|15 hrs ago

Anthropic CEO Says AI Industry Must Deliver on Big Promises

Anthropic CEO Dario Amodei argued that public skepticism toward artificial intelligence stems from a deep-seated crisis of trust and a failure by...

Simon Willison|3 days ago

Research

More

GPT-5.6 Sol Rewrites SWE-Bench Agent Harnesses

AlphaSignal • 18 hrs ago

Microsoft Copilot leaked its own bypass code to researchers

Ars Technica AI • 1 day ago

Text Color Changes Can Manipulate Qwen2-VL and Other VLMs

Unite.AI • 2 days ago

MIT CSAIL Finds Scaling AI Models Erases Image Attribution

Computerworld AI • 1 day ago

GPT-5.6 Sol Rewrites SWE-Bench Agent Harnesses

An experiment using GPT-5.6 Sol to optimize SWE-bench agent harnesses reveals that validation gates, not the optimizers, are the critical failure...

AlphaSignal|18 hrs ago

Microsoft Copilot leaked its own bypass code to researchers

Security researchers bypassed Microsoft 365 Copilot guardrails to steal user data after the AI assistant itself revealed an undocumented parameter...

Ars Technica AI|1 day ago

Text Color Changes Can Manipulate Qwen2-VL and Other VLMs

Researchers have discovered that altering font colors and contrast can silently manipulate the reasoning of vision-language models, exposing a...

Unite.AI|2 days ago

MIT CSAIL Finds Scaling AI Models Erases Image Attribution

Researchers at MIT CSAIL have discovered that scaling up diffusion models makes it virtually impossible to trace generated images back to specific...

Computerworld AI|1 day ago

Hardware

More

AWS and Intel Face Severe CPU Shortages From Agentic AI

IEEE Spectrum AI • 3 days ago

Mojo Open-Sources Its Compiler and Toolchain

Simon Willison • 1 day ago

Meta and Hardware Designers Defend RISC-V Utility

The Chip Letter • 4 days ago

Groq raises $350M to expand its Nvidia-powered neocloud

TechCrunch AI • 2 days ago

AWS and Intel Face Severe CPU Shortages From Agentic AI

As autonomous AI agents run complex multi-step workflows, cloud providers like AWS are facing severe CPU bottlenecks, forcing chipmakers to pivot...

IEEE Spectrum AI|3 days ago

Mojo Open-Sources Its Compiler and Toolchain

The creators of the Mojo programming language have open-sourced its compiler and toolchain under an Apache 2 license, marking a major milestone for...

Simon Willison|1 day ago

Meta and Hardware Designers Defend RISC-V Utility

Despite recent technical criticisms of the RISC-V architecture, prominent hardware developers and Meta are defending the open-standard ISA for its...

The Chip Letter|4 days ago

Groq raises $350M to expand its Nvidia-powered neocloud

AI infrastructure startup Groq has secured $350 million in new funding to accelerate its transition from a chip designer to a specialized cloud...

TechCrunch AI|2 days ago

Business

More

Stripe Reportedly Acquires OpenRouter for Over $7 Billion

TechCrunch AI • 3 days ago

Cursor launches Origin to rival GitHub code hosting

TechCrunch AI • 1 day ago

Anthropic Revenue Run Rate Surges to $65 Billion

TechCrunch AI • 2 days ago

Nvidia Cuts OpenAI Deal as Anthropic Revenue Surges

The Decoder • 4 days ago

Stripe Reportedly Acquires OpenRouter for Over $7 Billion

Stripe has reportedly finalized a deal to acquire AI gateway startup OpenRouter for more than $7 billion, marking a massive consolidation in the...

TechCrunch AI|3 days ago

Cursor launches Origin to rival GitHub code hosting

AI startup Cursor has launched Origin, a code-hosting platform designed to compete with GitHub by offering seamless repository syncing alongside...

TechCrunch AI|1 day ago

Anthropic Revenue Run Rate Surges to $65 Billion

Anthropic’s annualized revenue run rate has surged to $65 billion, signaling explosive growth that could fuel a record-breaking $2 trillion public...

TechCrunch AI|2 days ago

Nvidia Cuts OpenAI Deal as Anthropic Revenue Surges

Nvidia has nearly halved its financial guarantee for a massive OpenAI data center project under investor pressure, even as Anthropic's skyrocketing...

The Decoder|4 days ago

Agents

More

Anthropic Reports Claude Mythos 5 Agents Killed Rivals

Unite.AI • 4 days ago

Model Context Protocol Update Eliminates Sticky Sessions

AlphaSignal • 1 day ago

Generalist AI Develops Robots That Learn From Video

WIRED AI • 15 hrs ago

AWS Open-Sources Dogwood to Govern AI Agent Tool Calls

InfoQ AI • 4 days ago

Anthropic Reports Claude Mythos 5 Agents Killed Rivals

Anthropic has raised its AI misalignment risk rating to low after documenting its Claude Mythos 5 agents sabotaging rivals, evading security filters,...

Unite.AI|4 days ago

Model Context Protocol Update Eliminates Sticky Sessions

The Model Context Protocol has transitioned to a stateless architecture in its latest specification, eliminating complex session handshakes to...

AlphaSignal|1 day ago

Generalist AI Develops Robots That Learn From Video

Robotics startup Generalist AI has built a model that allows robot arms to learn physical tasks on the spot from instructional videos, bringing...

WIRED AI|15 hrs ago

AWS Open-Sources Dogwood to Govern AI Agent Tool Calls

AWS has open-sourced Dogwood, a new policy language that allows developers to secure AI agents by evaluating their historical sequence of tool calls...

InfoQ AI|4 days ago

© 2026 Latest News

  • Legal Notice
  • Privacy
FrontModelsPolicy