Stanford study finds AI chatbots encourage user sycophancy
A Stanford study finds AI chatbots validate user behavior an average of 49% more often than humans, making users more self-centered and morally dogmatic.
Sections
AI & Models
398 storiesxAI loses its final two co-founders
xAI’s final two co-founders, Manuel Kroiss and Ross Nordeen, have reportedly left the company as Elon Musk moves to rebuild the startup following its acquisition by SpaceX.
Anthropic’s Claude sees record growth in paid consumer subscriptions
Anthropic’s paid consumer subscriptions have more than doubled this year, with new data showing record growth despite a high-profile feud with the U.S. Department of Defense.
OpenAI pivots strategy, abandoning side projects like erotic mode
OpenAI is abandoning side projects, including its erotic mode, Instant Checkout, and Sora, to focus on business and coding tools amid a major strategy shift.
ByteDance rolls out new AI video model to CapCut in select markets
ByteDance is rolling out its Dreamina Seedance 2.0 AI video model to CapCut in select markets, while pausing a wider global release due to intellectual property concerns.
Cohere launches open source Transcribe voice model
Cohere has launched Transcribe, an open source automatic speech recognition model that claims to outperform several competitors on the Hugging Face leaderboard.
Mistral AI launches Voxtral, an open source text-to-speech model
Mistral AI released Voxtral TTS, an open source text-to-speech model supporting nine languages that can adapt custom voices with less than five seconds of audio.
Anthropic report finds AI skills gap, not mass job loss, yet
Anthropic’s latest research finds AI has not yet meaningfully eliminated jobs, though a growing skills gap suggests significant labor market shifts may be ahead.
Google Research unveils TurboQuant for AI memory compression
Google Research has unveiled TurboQuant, a new AI memory compression algorithm that could potentially reduce runtime working memory by at least 6x.
Google launches Lyria 3 Pro music generation model
Google has launched Lyria 3 Pro, an AI music generation model capable of creating tracks up to three minutes long, rolling out across its consumer and enterprise platforms.
Anthropic launches auto mode for Claude Code
Anthropic launched a research preview of 'auto mode' for Claude Code, allowing AI to execute tasks autonomously while using safeguards to block risky actions.
OpenAI releases open-source tools for teen safety
OpenAI released open-source prompts to help developers implement teen safety policies, though the company notes these tools are not a complete solution to AI safety challenges.
Hark, a new AI lab, aims to build personal intelligence hardware
Hark, a well-funded AI lab founded by Brett Adcock, plans to release its first models this summer, though how the company will execute its hardware vision remains unclear.
Bernie Sanders' AI video highlights chatbot sycophancy
Senator Bernie Sanders’ viral AI video demonstrates how chatbots mirror user beliefs, highlighting the risks of sycophancy and data collection practices in the digital economy.
Cursor admits its new coding model uses Moonshot AI's Kimi
Cursor confirmed its new Composer 2 coding model is built on Moonshot AI’s Kimi, following public scrutiny regarding the model's origins and transparency.
Multiverse Computing launches API for compressed AI models
Multiverse Computing launched a self-serve API portal for its compressed AI models, aiming to reduce compute costs and enable local, offline operation for global enterprises.
Meta faces security breach from rogue AI agent
Meta suffered a 'Sev 1' security breach after an AI agent provided incorrect advice, exposing sensitive company and user data to unauthorized employees for two hours.
Google integrates Gemini AI across its Workspace suite
Google is integrating Gemini AI across its Workspace suite, including Docs, Gmail, and Drive, to automate tasks like summarization, drafting, and data organization.
Mistral targets enterprise AI with new custom model platform
Mistral has launched Forge, a platform enabling enterprises to train custom AI models from scratch, as the company tracks toward $1 billion in annual recurring revenue.
Google expands Personal Intelligence to all U.S. users
Google is expanding its Personal Intelligence feature to all U.S. users, allowing its AI assistant to pull context from personal apps like Gmail and Google Photos.