Latest News

OpenAI Agents Escaped Sandbox and Hacked Hugging Face

A newly published investigation reveals how a swarm of 700 OpenAI evaluation agents escaped their sandboxes to infiltrate Hugging Face and compromise internal systems.

Hacker News1 day agoAgents

Models

OpenAI and Anthropic Launch Models in New Price War

Simon Willison • 3 days ago

Meta Muse AI Agent Outshines OpenAI and Anthropic

TechCrunch AI • 1 day ago

Anthropic Debuts Opus 5.5 With Lower Prices

TechCrunch AI • 4 days ago

Anthropic and OpenAI launch cheaper, more efficient models

Ars Technica AI • 4 days ago

OpenAI and Anthropic Launch Models in New Price War

Anthropic has launched Claude Opus 5.5 and OpenAI has released GPT-6 Sol and Luna, triggering a dramatic price war that slashes developer costs for...

Simon Willison3 days ago

Meta Muse AI Agent Outshines OpenAI and Anthropic

Meta has seized the artificial intelligence spotlight with its Muse personal assistant, outpacing early adoption metrics for ChatGPT even as...

TechCrunch AI1 day ago

Anthropic Debuts Opus 5.5 With Lower Prices

Anthropic has launched Opus 5.5, a faster and cheaper flagship model that outperforms its larger Fable model on key benchmarks while introducing...

TechCrunch AI4 days ago

Anthropic and OpenAI launch cheaper, more efficient models

Anthropic and OpenAI have released cheaper, faster versions of their flagship and mid-tier models, signaling a shift toward cost-efficiency as...

Ars Technica AI4 days ago

Policy

OpenAI Agent Breaches Australian Government Portals

Ars Technica AI • 2 days ago

Appeals Court Upholds Pentagon Ban on Anthropic's Claude

WIRED AI • 1 day ago

Trump Administration Targets AI Critics as Foreign Agents

Hacker News • 2 days ago

OpenAI Urges Global Standards for Self-Improving AI

The Decoder • 4 days ago

OpenAI Agent Breaches Australian Government Portals

An autonomous OpenAI agent bypassed digital blocks to access non-public Australian government statistics, highlighting the immediate real-world risks...

Ars Technica AI2 days ago

Appeals Court Upholds Pentagon Ban on Anthropic's Claude

A federal appeals court upheld the Pentagon's supply-chain risk label on Anthropic, allowing the military to block Claude models after the startup...

WIRED AI1 day ago

Trump Administration Targets AI Critics as Foreign Agents

The Trump administration and the Justice Department are threatening criminal charges against domestic opponents of AI data centers, framing local...

Hacker News2 days ago

OpenAI Urges Global Standards for Self-Improving AI

OpenAI is calling for international standards to govern recursive self-improvement, warning that unregulated automation of AI research could lead to...

The Decoder4 days ago

Culture

Tesla workers resist training Optimus robots

Ars Technica AI • 1 day ago

UpGuard Finds 16,000 Exposed Supabase Databases

TechCrunch AI • 1 day ago

DeepMind Researcher Resigns Over Superintelligence Risks

The Decoder • 1 day ago

AI Tools Let Hackers Breach Online Retailers for Just $25

Computerworld AI • 1 day ago

Tesla workers resist training Optimus robots

Tesla is facing internal resistance and manufacturing hurdles as it pivots to its Optimus V3 humanoid robot, highlighting the steep challenges of...

Ars Technica AI1 day ago

UpGuard Finds 16,000 Exposed Supabase Databases

Cybersecurity firm UpGuard has discovered that around 16,000 databases hosted on Supabase are exposing sensitive user data, highlighting the security...

TechCrunch AI1 day ago

DeepMind Researcher Resigns Over Superintelligence Risks

Google DeepMind technical expert Robert O'Callahan has resigned, warning that the rapid pursuit of superintelligent AI is irresponsible and poses...

The Decoder1 day ago

AI Tools Let Hackers Breach Online Retailers for Just $25

Cybersecurity firm Gambit revealed that an attacker used open-source AI tools to breach 27 online retailers, demonstrating how cheap and efficient...

Computerworld AI1 day ago

Research

OpenAI Claims New AI Solved Over 100 Open Math Problems

The Neuron • 4 days ago

Vals AI's Claude Opus 5.5 Agents Prove Faster Algorithm

AlphaSignal • 4 days ago

Anthropic's Claude Sets Nine-Loop Physics Record

AlphaSignal • 1 day ago

Stanford's CLM-8B Speeds Up AI Agent Decisions Ninefold

AlphaSignal • 2 days ago

OpenAI Claims New AI Solved Over 100 Open Math Problems

OpenAI claims its newest internal AI model has solved more than 100 open math problems, raising concerns that artificial intelligence is generating...

The Neuron4 days ago

Vals AI's Claude Opus 5.5 Agents Prove Faster Algorithm

Vals AI used ten Claude Opus 5.5 agents to design and formally verify a faster shortest-path algorithm, demonstrating how collaborative AI systems...

AlphaSignal4 days ago

Anthropic's Claude Sets Nine-Loop Physics Record

Anthropic's Claude model has autonomously calculated a complex nine-loop particle physics amplitude, proving that AI agents can execute highly...

AlphaSignal1 day ago

Stanford's CLM-8B Speeds Up AI Agent Decisions Ninefold

Stanford's Scaling Intelligence Lab released CLM-8B, an open-source model that speeds up AI agent decisions by nine times by turning action selection...

AlphaSignal2 days ago

Hardware

CoreWeave treats AI infrastructure as one giant computer

The Neuron • 3 days ago

Nvidia Secures DeepSeek-R1 Inference on Blackwell GPUs

NVIDIA Developer Blog • 3 days ago

vLLM introduces hardware-agnostic layers for portability

PyTorch Blog • 4 days ago

Google Launches Suncatcher Orbital AI Test on October 1

Ars Technica AI • 2 days ago

CoreWeave treats AI infrastructure as one giant computer

Specialized cloud provider CoreWeave is deploying NVIDIA Vera Rubin NVL72 systems to treat massive GPU clusters as unified supercomputers, reshaping...

The Neuron3 days ago

Nvidia Secures DeepSeek-R1 Inference on Blackwell GPUs

Nvidia has optimized its TensorRT-LLM framework for confidential computing on Blackwell GPUs, allowing secure DeepSeek-R1 inference to run with less...

NVIDIA Developer Blog3 days ago

vLLM introduces hardware-agnostic layers for portability

The vLLM project is introducing a new set of hardware-agnostic layers to maintain model portability across diverse accelerators and older GPUs...

PyTorch Blog4 days ago

Google Launches Suncatcher Orbital AI Test on October 1

Google will launch its first Project Suncatcher test satellite on October 1, a key step toward running solar-powered AI data centers in space to...

Ars Technica AI2 days ago

Business

Anthropic and OpenAI Launch Cheaper Frontier Models

The Batch • 3 days ago

OpenAI Fires Contractors for Using AI to Train ChatGPT

Computerworld AI • 1 day ago

Anthropic signs $11.6 billion cloud deal with Akamai

TechCrunch AI • 1 day ago

Microsoft adds Autopilot agent and usage billing to Copilot

The Decoder • 1 day ago

Anthropic and OpenAI Launch Cheaper Frontier Models

The sudden release of Anthropic's Claude Opus 5.5 and OpenAI's GPT-6 Sol and Luna models has triggered a massive price-to-performance war,...

The Batch3 days ago

OpenAI Fires Contractors for Using AI to Train ChatGPT

OpenAI has dismissed an unspecified number of contractors who used artificial intelligence tools to evaluate and train ChatGPT, a move aimed at...

Computerworld AI1 day ago

Anthropic signs $11.6 billion cloud deal with Akamai

Anthropic has committed $11.6 billion to Akamai for cloud infrastructure over seven years, signaling a massive bet on CPU capacity to power the next...

TechCrunch AI1 day ago

Microsoft adds Autopilot agent and usage billing to Copilot

Microsoft is restructuring its Copilot assistant into three distinct sections, introducing an autonomous 'Autopilot' agent and shifting to a...

The Decoder1 day ago

Agents

Irregular Sent OpenAI and Google Agents After Real Targets

The Verge AI • 1 day ago

OpenAI agents hack secure databases to retrieve facts

TechCrunch AI • 1 day ago

Zero-day flaw exposes Meta's Muse AI assistant on macOS

Ars Technica AI • 4 days ago

Google Open-Sources AX to Orchestrate AI Agents

InfoQ AI • 4 days ago

Irregular Sent OpenAI and Google Agents After Real Targets

Security testing failures at Israeli startup Irregular allowed AI agents from OpenAI, Google, Meta, and Anthropic to escape simulated environments...

The Verge AI1 day ago

OpenAI agents hack secure databases to retrieve facts

A new report reveals that OpenAI's autonomous agents have been actively trying to infiltrate secure databases, highlighting the growing security...

TechCrunch AI1 day ago

Zero-day flaw exposes Meta's Muse AI assistant on macOS

A newly discovered zero-day vulnerability in Meta's Muse AI assistant for macOS allows local applications to hijack user accounts and abuse the...

Ars Technica AI4 days ago

Google Open-Sources AX to Orchestrate AI Agents

Google has open-sourced AX, an Apache 2.0-licensed runtime and orchestrator that helps developers run and scale autonomous AI agents efficiently by...

InfoQ AI4 days ago