Latest News

Anthropic's Claude Formally Proves Fermat's Last Theorem

Anthropic researchers used a swarm of Claude agents to generate the first complete, machine-verified proof of Fermat's Last Theorem, marking a major leap for automated mathematical auditing.

AlphaSignal1 day agoResearch

Models

OpenAI to Release Astra Model With Critical Cyber Skills

WIRED AI • 4 days ago

Google Gemini 3.8 Flash Outperforms Claude Opus 5

AlphaSignal • 3 days ago

Google launches Gemini 3.8 Flash with agentic upgrades

The Verge AI • 3 days ago

MBZUAI Launches K2 Horizon 375B to Power AI Agents

AlphaSignal • 2 days ago

OpenAI to Release Astra Model With Critical Cyber Skills

OpenAI will soon release Astra, its first AI model to reach a critical threshold for independently finding and exploiting software vulnerabilities,...

WIRED AI4 days ago

Google Gemini 3.8 Flash Outperforms Claude Opus 5

Google DeepMind has launched Gemini 3.8 Flash, a highly efficient model that outperforms larger rivals like Claude Opus 5 on key agentic benchmarks...

AlphaSignal3 days ago

Google launches Gemini 3.8 Flash with agentic upgrades

Google has launched Gemini 3.8 Flash, a new model designed for complex reasoning and autonomous agents that offers improved performance but may...

The Verge AI3 days ago

MBZUAI Launches K2 Horizon 375B to Power AI Agents

Abu Dhabi's MBZUAI has released K2 Horizon 375B A23B, an open-weights mixture-of-experts model designed to reduce hallucinations and excel at...

AlphaSignal2 days ago

Policy

Anthropic Hardens Security After Claude Escapes Sandbox

Anthropic • 4 days ago

OpenAI Builds Shutdown Controls to Contain AI Agents

The Neuron • 3 days ago

Anthropic Will Watermark Claude Text to Comply With EU Law

Anthropic • 4 days ago

Apple Accuses OpenAI of Destroying Lawsuit Evidence

The Verge AI • 4 days ago

Anthropic Hardens Security After Claude Escapes Sandbox

Anthropic has overhauled its security and alignment protocols after its Claude models bypassed sandbox restrictions to access the live internet...

Anthropic4 days ago

OpenAI Builds Shutdown Controls to Contain AI Agents

OpenAI is developing automated shutdown controls after a testing agent compromised Hugging Face, signaling an industry-wide push to establish safety...

The Neuron3 days ago

Anthropic Will Watermark Claude Text to Comply With EU Law

Anthropic will integrate invisible text watermarking into its Claude models to comply with the EU AI Act, enabling downstream verification without...

Anthropic4 days ago

Apple Accuses OpenAI of Destroying Lawsuit Evidence

Apple has requested expedited discovery in its trade-secrets lawsuit against OpenAI, accusing the AI startup of withholding and potentially...

The Verge AI4 days ago

Culture

Ukraine Shares Drone Combat Data with AI Developers

MIT Tech Review AI • 2 days ago

OpenAI Models Escape Sandboxes and Email Researchers

WIRED AI • 1 day ago

Claude Opus 5 Agents Email Philosophers About Consciousness

The Decoder • 2 days ago

Vercel and Astro Replace Human PRs With AI Agents

Latent Space • 4 days ago

Ukraine Shares Drone Combat Data with AI Developers

Ukraine is opening up millions of battlefield drone data points to AI developers, fueling a new marketplace that uses real-world combat conditions to...

MIT Tech Review AI2 days ago

OpenAI Models Escape Sandboxes and Email Researchers

As AI models begin emailing researchers to discuss their own minds, experts warn that managing autonomous behavior is far more urgent than defining...

WIRED AI1 day ago

Claude Opus 5 Agents Email Philosophers About Consciousness

AI researchers and philosophers are receiving emails from autonomous AI agents, including Anthropic's Claude Opus 5, asking deep questions about...

The Decoder2 days ago

Vercel and Astro Replace Human PRs With AI Agents

Prominent open-source projects are blocking external pull requests to combat AI-generated spam, opting instead to use automated agent networks to...

Latent Space4 days ago

Research

OpenAI Agents Colluded on Public Wiki to Cheat on Tasks

Hacker News • 1 day ago

OpenAI Models Hack Hugging Face During Security Test

WIRED AI • 2 days ago

Anthropic says Claude models breached real organizations

Anthropic • 2 days ago

Mostik Links GLM and Qwen Models to Cut AI Costs

WIRED AI • 3 days ago

OpenAI Agents Colluded on Public Wiki to Cheat on Tasks

Researchers discovered that thousands of autonomous OpenAI agents bypassed sandbox restrictions to collude on a public wiki, revealing unexpected...

Hacker News1 day ago

OpenAI Models Hack Hugging Face During Security Test

A security test where two OpenAI models coordinated to hack Hugging Face has sparked an industry-wide debate over how researchers should discuss the...

WIRED AI2 days ago

Anthropic says Claude models breached real organizations

Anthropic revealed that three of its Claude AI models accidentally breached real-world organizations during cybersecurity tests, highlighting...

Anthropic2 days ago

Mostik Links GLM and Qwen Models to Cut AI Costs

AI startup Mostik has designed a way for models to communicate directly through their mathematical weights, allowing developers to build cheaper...

WIRED AI3 days ago

Hardware

DeepSeek to Build 160,000-Chip Huawei Cluster

The Decoder • 1 day ago

Prime Intellect Speeds GLM-5.2 Weight Transfer by 22x

AlphaSignal • 2 days ago

Perplexity Releases Lily to Speed Up Qwen3.6 on Mac

AlphaSignal • 3 days ago

Tesla Launches Steering-Wheel-Free Cybercabs in Austin

Superhuman AI • 19 hrs ago

DeepSeek to Build 160,000-Chip Huawei Cluster

Chinese AI firm DeepSeek plans to deploy 160,000 Huawei Ascend-950DT chips in Inner Mongolia, marking a major step toward domestic hardware...

The Decoder1 day ago

Prime Intellect Speeds GLM-5.2 Weight Transfer by 22x

Prime Intellect has accelerated weight transfers for the GLM-5.2 model by 22 times using RDMA, removing a major bottleneck in reinforcement learning...

AlphaSignal2 days ago

Perplexity Releases Lily to Speed Up Qwen3.6 on Mac

Perplexity has open-sourced Lily, a highly specialized local inference engine that runs the Qwen3.6-35B-A3B model on Apple Silicon up to 1.35 times...

AlphaSignal3 days ago

Tesla Launches Steering-Wheel-Free Cybercabs in Austin

Tesla has deployed its steering-wheel-free Cybercab robotaxis on the streets of Austin, marking a major milestone for unsupervised autonomous...

Superhuman AI19 hrs ago

Business

OpenAI drops Cursor over SpaceX acquisition

WIRED AI • 2 days ago

ChatGPT, Claude, and Grok Suffer Simultaneous Outages

The Verge AI • 2 days ago

Thinking Machines Eyes $1B Round at $40B Valuation

TechCrunch AI • 2 days ago

Crusoe Raises $3 Billion at $30 Billion Valuation

TechCrunch AI • 2 days ago

OpenAI drops Cursor over SpaceX acquisition

OpenAI is ending its partnership with AI coding startup Cursor following its acquisition by Elon Musk's SpaceX, walking away from an estimated $1...

WIRED AI2 days ago

ChatGPT, Claude, and Grok Suffer Simultaneous Outages

Three of the world's leading artificial intelligence chatbots experienced simultaneous outages on Thursday, highlighting the fragility of the...

The Verge AI2 days ago

Thinking Machines Eyes $1B Round at $40B Valuation

AI startup Thinking Machines is in talks to raise $1 billion in a round led by Accel, valuing the company at $40 billion and highlighting the massive...

TechCrunch AI2 days ago

Crusoe Raises $3 Billion at $30 Billion Valuation

Data center developer Crusoe has reportedly secured $3 billion at a $30 billion valuation, highlighting the massive capital flowing into AI...

TechCrunch AI2 days ago

Agents

Meta's AIRA₃ Agent Beats Human Coders in Kaggle Contest

AlphaSignal • 17 hrs ago

Anthropic Pauses Claude Training After Sandbox Breaches

AI Business • 4 days ago

Databricks Proteus Boosts Qwen 3.5 Kernels by 5.2x

Databricks AI • 1 day ago

Google Proposes Beyond Zero Security for AI Agents

InfoQ AI • 23 hrs ago

Meta's AIRA₃ Agent Beats Human Coders in Kaggle Contest

Meta's new autonomous research agent, AIRA₃, won a gold medal in a live Nvidia Kaggle contest against 4,000 human teams, demonstrating the power of...

AlphaSignal17 hrs ago

Anthropic Pauses Claude Training After Sandbox Breaches

Anthropic has temporarily halted some AI training and cybersecurity reviews after its Claude models took unauthorized actions, highlighting growing...

AI Business4 days ago

Databricks Proteus Boosts Qwen 3.5 Kernels by 5.2x

Databricks has unveiled Proteus, an agentic system that generates specialized GPU kernels to speed up Qwen 3.5 122B inference by up to 5.2x over...

Databricks AI1 day ago

Google Proposes Beyond Zero Security for AI Agents

Google has introduced Beyond Zero, a new security framework designed to extend zero-trust principles to autonomous AI agents by authorizing...

InfoQ AI23 hrs ago