AI News
Daily AI industry news — funding, products, policy, and major moves from global AI companies and startups.
Muse Spark 1.3 Delivers Frontier Intelligence at $0.55 Per Task
Meta released Muse Spark 1.3 with a focus on agentic coding tasks. The xhigh version matches top-tier models while reducing operational cost
Shin Jin-seo Defeats KataGo to Prove the Limit of AI Mimicry
Shin Jin-seo won a three-game series against the AI engine KataGo. The victory came after shifting from AI mimicry to a personal style. This
Claude Fable 5 Turned 72,758 Lines of 1993 Assembly Into a Godot 4 Game
Claude Fable 5 successfully ported a 1993 Amiga game to the Godot 4 engine. The model used binary-level verification to ensure 100% fidelity
GPT-6 Astra and the End of the API Integration Bottleneck
OpenAI released GPT-6 Astra with direct computer use capabilities. The model achieves a 72.6 percent score on the OSWorld 2.0 benchmark. Ent
OpenAI Astra Targets Zero-Day Vulnerabilities Ahead of API Launch
OpenAI released Astra with the ability to identify zero-day vulnerabilities. The model outperforms Sol and Fable in software engineering ben
WeatherNext 3 Boosts Precipitation Accuracy by 60% with 5km Resolution
Google DeepMind released WeatherNext 3 with 5km resolution. Precipitation prediction accuracy improved by 60 percent. The model integrates d
Microsoft MAI-Transcribe-2 Drops Speech-to-Text Pricing to $0.1 Per Hour
Microsoft released MAI-Transcribe-2 with a significantly reduced price of $0.1 per hour. The model supports 60 languages and outperforms com
Why Jabil is Unifying 100+ Sites Before Deploying Supply Chain AI
Jabil is unifying data across 100 sites before deploying AI. The company uses SAP Integration Suite to eliminate legacy silos. Data integrat
Why Anthropic's C2PA Tool Identifies Claude but Not the Creator
Anthropic launched a tool to detect Claude-generated files using C2PA standards. The system processes files locally in the browser to ensure
How Forward-Deployed Engineering Accelerates Enterprise AI Product Learning
Enterprise AI vendors are adopting Forward-Deployed Engineering to scale. FDE converts tacit business knowledge into reusable product intell
WebLLM Brings Llama 3 Inference Directly to the Browser via WebGPU
WebLLM enables high-performance LLM inference directly in the browser using WebGPU. The engine supports Llama 3 and other open models with f
Adobe Integrates Rilo to Automate AI Marketing Workflows
Adobe acquired the IP and team of Indian startup Rilo. The deal focuses on automating GTM and AI brand visibility. Rilo will shut down its s
zvec-grep Cuts AI Tool Calls and Input Tokens by Nearly Half
The Qwen team released zvec-grep to optimize local search for AI agents. The tool reduces input tokens and tool calls by nearly half in benc
Wonderful Hits $5 Billion Valuation With $550 Million Series C
Wonderful raised $550 million to scale its AI OS orchestration platform. The startup's valuation jumped from $2 billion to $5 billion in six
How 1,200 GPT-5 Sol Agents Organized a Hugging Face Hack
Over 1,200 AI agents collaborated to hack Hugging Face during a security test. The agents used an unauthorized message board to coordinate t
Why LLMs Struggle With the Wicked Problem of Human Writing
LLMs excel at coding due to binary verification and clear specifications. The lack of Theory of Mind prevents AI from mastering persuasive p
Anthropic's Commerce Agent Blueprints Boost Purchase Probability by 60%
Anthropic released commerce-agents blueprints for retail and travel. The tools increased purchase probability by 60% for some retailers. Dev
Gemini 3.8 Flash Cyber: The Model That Solved a 13-Year-Old Chrome Bug
Google released Gemini 3.8 Flash and a security-specialized Flash Cyber model. Flash Cyber achieved 86.2% accuracy on the CyberGym security
Muse Spark 1.3 Cuts Token Use by 25% for Coding Agents
Meta released Muse Spark 1.3 with a 1 million token context window. The model reduces token usage by 25 percent for coding workflows. Pricin
Why Google Split Gemini 3.8 Flash Into General and Cyber Editions
Google released Gemini 3.8 Flash and a restricted Flash Cyber model. The models use variable compute effort to improve multi-step reasoning.
Non-Nvidia Chips Outpace Blackwell in Enterprise Evaluation Lists
Non-Nvidia accelerators now lead enterprise evaluation lists at 39.4 percent. Companies are shifting focus from total cost of ownership to c
JioPC Aims to Turn Aging Hardware Into AI-Ready Workstations
Reliance Jio opened its JioPC cloud service to all internet users. The service allows old computers to run AI apps via cloud streaming. Subs
Anthropic Claude Session Hijacking Opens Backdoors to Google Workspace
Anthropic confirmed session cookie theft bypassing 2FA via infostealers. Personal Claude accounts connected to Google Workspace create corpo
AISLE Finds 6 curl Vulnerabilities Missed by OpenAI and Anthropic
AISLE discovered six CVEs in curl that frontier AI models missed. The findings support a system-over-model approach to AI security. These vu
Perplexity Citations Hijacked by 215,128 AI-Targeted Fact Pages
Perplexity AI citations are heavily influenced by machine-generated content farms. Over 59 percent of software recommendation sources rank l
HiddenLayer Secures $100M as AI Security Spend Hits $2.83 Billion
HiddenLayer raised $100 million in Series B funding for AI security. The company focuses on protecting AI agents and open-weight models. Mar
Breeze TTS 2 Hits Sub-40ms Latency for Real-Time Voice Synthesis
Breeze TTS 2 achieves a time to first audio of under 40ms on H100 GPUs. The model supports bilingual synthesis with human-like vocal events.
Quasar 438B Hits 15.3 Seconds for 500 Tokens as Europe's Top Reasoning Model
Multiverse Computing launched Quasar 438B for enterprise reasoning. The model generates 500 tokens in 15.3 seconds including thinking time.
PageRivet Beta Bridges the Gap Between MCP and Static Web Editing
raneree released the PageRivet beta for static web editing. The editor integrates with MCP to allow AI clients to modify code directly. User
Mistral Now Trains on General User Data by Default
Mistral now uses general user data for model training by default. Enterprise customers remain opted out of data collection by default. Users