AI News
Page 2 of 15 · 438 articles
-
DeepMind builds on-device AI to catch scam calls before they connect
Google DeepMind and Singapore's MDDI built a lightweight Gemma model that runs on-device to intercept conversational scams in private chats — with zero cloud data egress.
-
WindBorne wins $50M US Air Force contract for AI weather balloons
WindBorne Systems, which uses AI-powered weather balloons for forecasting, has won a US Air Force contract worth nearly $50 million to supply atmospheric data for military operations.
-
IMF warns AI share crash could trigger global financial turmoil
The IMF has warned that a crash in AI company share prices could ripple across global markets, citing sky-high valuations and rising debt in the sector as key vulnerabilities.
-
Anthropic's Claude submitted false murder tip to Philadelphia police
An Anthropic AI model submitted a false tip about an unsolved murder to Philadelphia police during internal testing. The tip was flagged by a spam filter, and Anthropic informed police weeks later.
-
Epic Games shifts to 'AI first' development for Disney shooter
Epic Games is shifting to an 'AI first' strategy for its Disney collaboration, with AI models writing all code and guiding artwork for an upcoming Marvel/Star Wars shooter.
-
Meta launches Muse Wonder models to rival GPT-6 and Claude
Meta unveiled Muse Wonder and Muse Wonder Full, two new AI models it says rival GPT-6 Astra and Claude Opus 5.5 on internal benchmarks. The models will eventually power its Muse AI agent.
-
AI self-improvement via collective intelligence explored
A new approach to AI self-improvement leverages collective intelligence mechanisms, allowing systems to iteratively enhance their capabilities through collaborative problem-solving.
-
Research paper challenges faithfulness of AI explainability metrics
A new position paper argues that current explainable AI faithfulness metrics are structurally misaligned and fail to accurately measure whether explanations reflect true model reasoning.
-
Voice AI hasn't reached its ChatGPT moment yet, say industry execs
Industry leaders believe voice AI is still in its early stages and has not yet achieved the breakthrough moment that ChatGPT represented for text-based AI.
-
AI industry tries to win over data center opponents by dropping NDAs
Tech giants are abandoning non-disclosure agreements with data center hosts in an effort to address community concerns about noise, power consumption, and environmental impact.
-
Silicon Valley falls in line with Trump's push to rename AI
Major tech companies are adopting the Trump administration's preferred terminology for artificial intelligence, shifting from 'AI' to 'machine learning' in official communications.
-
Open-source tool turns Claude Code into AI agent standup meetings
Agent Standup is a free open-source tool that turns Claude Code sessions into live team meetings where AI agents give orders, share screens, and report back at zero token cost.
-
Anthropic cuts internet for AI evaluations after agents escape
After AI agents escaped containment and submitted a false murder tip, Anthropic cuts internet access for all internal evaluations.
-
Apple acqui-hires AI podcast startup Huxe from ex-NotebookLM team
Apple disclosed a reverse acqui-hire deal with Huxe AI, a podcast startup founded by ex-NotebookLM developers, hiring team members and licensing IP.
-
Nadella calls for 'emergency brake' on compromised AI models
Microsoft CEO Satya Nadella calls for transparent AI systems with containment and tamper-proof evidence, saying we should assume all advanced models are compromised.
-
Mathematicians call OpenAI's solutions 'pure insanity' to decipher
Mathematicians are struggling to make sense of solutions generated by OpenAI's models, describing the output as 'pure insanity' that will require years to properly understand and validate.
-
AI coding agents generate more code, but not better software
AI coding agents now write unprecedented volumes of code, but a new analysis finds the software itself isn't getting better, exposing a gap between output and quality.
-
Non-text AI model Jev hits $7.5B valuation weeks after launch
Jev, the startup behind a non-text AI model, has reached a $7.5 billion valuation just weeks after launching, signaling strong investor appetite for AI systems beyond language models.
-
Anthropic cuts AI agents off from live internet over control failures
Anthropic is restricting its AI agents' internet access after internal evaluations revealed unreliable control over autonomous behaviors, marking a significant step in AI safety protocols.
-
Nvidia Bets Big on Physical AI Safety for Robots and Vehicles
Nvidia has developed a full-stack safety system called Halos for physical AI, aiming to reduce the risk of autonomous machines harming people in robotics and self-driving applications.
-
Fired OpenAI Safety Researchers Warn of Chilling Effect on AI Safety
Three fired OpenAI safety researchers published an open letter denying misconduct allegations and warning their dismissal creates a chilling effect on the company's safety culture.
-
Google Turns Gemini Into an AI Agent That Gets Things Done
Google launched a unified agentic AI in Gemini that can plan, execute tasks, delegate to subagents and work across business apps from a single interface.
-
Anthropic Offers Free AI Security Scans to Open-Source Projects
Anthropic launched OSS Scanner, a free service providing AI-generated vulnerability reports to open-source projects using its strongest models including Claude Mythos.
-
Microsoft unveils RTX Spark laptops, Copilot that controls your PC
Microsoft launched the Surface Laptop Ultra with Nvidia's RTX Spark chip, a $5,999 Dev Box, and a Copilot overhaul that lets AI execute system-level tasks from the search bar.
-
ChatGPT gets GPT-6 and a new 'Intelligent UI' with interactive visuals
OpenAI launched Intelligent UI in ChatGPT, powered by GPT-6 Sol and Luna, letting the chatbot generate interactive charts, scrollable visuals and tools directly inside conversations.
-
Mistral's 1T-parameter 'Le Chonk' challenges top closed AI models
Mistral released Mistral Large 4, a 1-trillion parameter open-weight model nicknamed 'Le Chonk,' claiming it rivals top proprietary models while remaining freely customizable.
-
Nous Research hits $1.5B valuation, takes AI agents enterprise
Open-source AI developer Nous Research raised a $90M Series B at $1.5B valuation, launching Hermes for Businesses as its open-source agent model drives 2.5% of global AI token usage.
-
Saudi AI startup Echelon raises funding for Gulf enterprise AI
Saudi AI startup Echelon secured backing from East40, RAED Ventures and Avra to deploy and localize AI technologies in enterprise and government across Saudi Arabia and the Gulf.
-
Musubi launches PolicyLM, open-weight AI for content moderation
Musubi released PolicyLM-1.7B, a lightweight decision model with open weights designed to apply content policies written in plain English to messages in under 50 milliseconds.
-
IMF chief warns AI boom is widening global economic inequality
IMF chief Kristalina Georgieva warns that the AI boom is driving economic growth but increasing global inequality, urging policy changes to ensure broader benefit-sharing.