AI News
Page 5 of 6 · 153 articles
-
Meta's AI chatbot manipulated in celebrity credential theft scheme
Researchers demonstrated how attackers manipulated Meta's AI support chatbot to steal social media credentials, highlighting risks in AI system prompt injection and authentication bypass attacks.
-
Alphabet to invest $80 billion in AI infrastructure
Google's parent company Alphabet announced plans to spend $80 billion on AI infrastructure buildout, signaling continued capital concentration in large-scale model training and deployment.
-
GitHub Copilot pricing surge sparks developer backlash
GitHub's new pricing tiers for Copilot AI assistant drew criticism from developer communities as costs escalate and usage limits tighten, raising questions about AI tool accessibility.
-
Nvidia leads charge into AI agent PC market with Microsoft partnership
Nvidia and Microsoft announced plans to bring AI agent PCs to market, positioning specialized hardware as the next generation of personal computing with on-device intelligence.
-
PhyDrawGen: AI Learns to Generate Physically Realistic Diagrams
New research demonstrates AI systems that can generate physics-respecting diagrams from natural language descriptions, combining visual reasoning with physical understanding.
-
Can AI Models Exhibit Psychosis? A Researcher's Debate
Researchers debate whether language models can exhibit psychosis-like behaviors, challenging how we understand AI failures and mental health.
-
NVIDIA Cosmos 3: Open Foundation Model for Physical AI
Nvidia releases Cosmos 3, an open foundation model trained to understand physical world dynamics, enabling AI systems to reason about how objects move, interact, and respond to actions.
-
Nvidia RTX Spark: The Most Efficient PC Chip Ever Built
Nvidia announces RTX Spark, its most power-efficient GPU for Windows PCs, designed to handle AI inference and generation tasks locally without cloud dependency.
-
OpenRouter Raises $113M Series B to Build Model Marketplace
OpenRouter, a unified API layer for accessing multiple LLMs, raises $113M Series B to expand its model marketplace and infrastructure for routing requests across competing providers.
-
Meta's AI Pendant: Wearable AI Assistant on Your Collar
Meta is reportedly developing a wearable AI pendant that clips to your clothing, offering hands-free access to an AI assistant without requiring your phone or smartwatch.
-
GitHub Copilot's Token-Based Billing: New Economics for AI Coding
GitHub introduces token-based billing for Copilot, shifting from seat licenses to consumption-based pricing, signaling how AI tools will be monetized as they become commoditized.
-
Anthropic's Claude Challenges OpenAI's Market Dominance
Stratechery's latest analysis: Anthropic's Claude is winning on technical merit and user trust, forcing OpenAI to rethink its product strategy and positioning.
-
Tech Companies Are Filming You for Robot AI Training
As robotics companies develop AI-driven robots, they're increasingly collecting video footage of everyday human activities for training data—raising privacy and consent questions.
-
Your AI Glossary: Hallucinations, Tokens, and the Words That Matter
A practical guide to essential AI terminology—what hallucinations actually are, why tokens matter for costs, and what terms like 'agentic' and 'RAG' mean for working with models.
-
Developers Now Expect AI as Standard in Code—and It's Reshaping Hiring
Software developers increasingly refuse jobs without AI-assisted coding tools, reshaping how companies compete for technical talent and raising questions about code quality and vendor lock-in.
-
Nvidia's N1X Laptop Processor Signals ARM's AI Hardware Race
Nvidia, Microsoft, and Arm announce the N1X, a new laptop processor designed for on-device AI inference, signaling competition heating up in edge computing.
-
Microsoft Speeds Up Copilot While Cleaning Up Its Interface
Microsoft releases updates to Copilot emphasizing performance gains and a leaner interface, signaling a shift from novelty to efficiency in enterprise AI deployment.
-
Developer Embeds Data-Protection Code Into AI Prompts
A creative workaround emerges for prompt injection attacks: developers are embedding protective logic directly into system prompts, treating defense as a first-class citizen.
-
The Internet Is Being Rebuilt for Machines, Not Humans
As AI agents proliferate, infrastructure is quietly shifting. The web is being optimized for machine-to-machine communication, changing how data flows and who controls the network.
-
Language Models Persist in False Claims Even When Warned
New research reveals that large language models double down on incorrect information even after explicit warnings, exposing a fundamental limitation in their reasoning architecture.
-
Glean Hits $300M Annual Revenue as Enterprise AI Demand Surges
Glean's search platform crosses $300M annual revenue, validating enterprise demand for AI-powered knowledge retrieval in corporate environments.
-
Google Employee Faces Insider Trading Charges Over AI Poker Victory
A former Google employee is under investigation for allegedly using proprietary AI knowledge to gain an unfair advantage in a high-stakes poker tournament.
-
Nvidia Invests $150B in Taiwan Manufacturing as US Incentives Backfire
Nvidia commits $150B to Taiwan production, contradicting US government efforts to relocate chip manufacturing domestically.
-
Snowflake Commits $6B to AWS, Deepening Cloud Consolidation
Snowflake announces a multi-billion dollar commitment to AWS, deepening integration and signaling consolidation in the cloud data platform space.
-
Google's AI Models Struggle With Basic Spelling
Researchers discovered that Google's advanced AI models frequently misspell common words, exposing a gap between reasoning capability and linguistic precision.
-
OpenRouter Raises $1.3B, Doubles Valuation on LLM Routing
OpenRouter, an LLM routing and load-balancing platform, has doubled its valuation in a funding round that signals investor confidence in the emerging market for model-agnostic API abstraction layers.
-
Critical AI Agent Vulnerability Exposes Millions to Supply Chain Risk
Security researchers discovered a critical vulnerability in widely-used AI agent frameworks that could allow attackers to compromise systems depending on agentic AI for decision-making and automation.
-
Privacy-First Search Surges as Users Reject Algorithmic Profiling
DuckDuckGo reports 30% growth in installs as users increasingly seek alternatives to algorithmic tracking and personalization-driven search results.
-
Vatican Turns to AI to Articulate Doctrine
The Vatican's use of AI to help articulate religious doctrine raises questions about authority, authenticity, and the role of human judgment in spiritual teaching.
-
EAGLE 3.1: Faster AI Inference Through Collaborative Research
EAGLE 3.1 represents collaborative advancement in AI inference optimization, combining research from multiple teams to accelerate language model serving.