DOWN TO BUSINESS

📊 Enterprise AI Adoption Stalls at Pilot Stage

KPMG's quarterly survey of 130 business leaders at large companies reveals that AI adoption is stalling at the pilot stage. While more businesses than ever are experimenting with AI technology, the percentage of companies deploying AI agents beyond pilot programs remains stagnant at just 11%. Traditional applications like customer support and data analysis dominate current enterprise use cases.

📋 Workday Using Customer Contracts for Training

Workday.com is using information from customer contracts to improve its AI-powered legal contract analysis tools. With customer consent, the HR software provider collects rental agreements and software sales documents to train models that generate new contracts and analyze potential risks.

🏫 US Government to Integrate AI into K-12 Education

The US government will integrate AI into K-12 classrooms, establish AI courses for students, and create a White House Task Force on AI Education. This initiative comes as the Department of Education faces restructuring, signaling a significant policy shift toward embracing AI technologies in primary and secondary education to prepare the next generation for an AI-driven workforce.

🚗 Volkswagen Builds Its Own Self-Driving Tech

Volkswagen just showed off its first self-driving system made fully in-house, built with its partner Horizon Robotics. Their new AI platform, GAIA, collects huge amounts of driving data and trains the software 20 times faster than before. The system will launch in a new Volkswagen electric car later this year, with more models coming in 2026.

📱 Netflix Tests AI Search Feature

Netflix is testing a new OpenAI-powered search feature that allows users to find shows based on mood or detailed prompts. Currently available to iOS users in Australia and New Zealand, with US expansion planned soon, this AI-driven approach shifts streaming from passive browsing to intent-driven discovery.

TRENDING TOPIC

🤖 OpenAI Developing Social Media Platform with "Yeeting”

OpenAI is reportedly working on its own social media platform, turning what seemed like a joke tweet from CEO Sam Altman into reality. Early prototypes center around ChatGPT's image generation capabilities, with posting to the feed currently known as "yeeting." The platform could provide OpenAI with valuable user data for training future AI models while creating a dedicated space for sharing AI-generated content.

HOT MODELS

🎨 Korean Students Create Top-Tier Voice Model with Zero Funding

Nari Labs, founded by two Korean undergraduate students with no prior experience, has released Dia, an open-source text-to-speech model that claims to outperform leading commercial offerings. Created in just three months with zero funding, the model supports advanced features like emotional tones, multiple speaker tags, and nonverbal cues like coughs and screams.

🔍 Anthropic Maps Claude's Moral Values

Anthropic has published a study analyzing hundreds of thousands of real AI conversations to understand how models like Claude make moral judgments. Researchers examined over 700,000 real but anonymized conversations to identify and categorize unique values expressed by the AI, creating the first large-scale map of an AI model's values in day-to-day interactions with users.

🎨 OpenAI's Image Generator Now Available for Developers

OpenAI has launched its advanced image generation model, GPT-Image-1, to developers via API, bringing the viral success of ChatGPT's image capabilities to third-party applications. The model shares the features that made GPT-4o's generations popular, including real-world knowledge and accurate text rendering, with companies like Adobe, Figma, Canva, and Instacart already incorporating it into their products.

💬 Anthropic to Release Voice Mode with Multiple Personalities

Reports suggest Anthropic is preparing to release a long-awaited voice mode for Claude featuring three distinct voice-based personalities named Airy, Mellow, and Buttery. It's designed to make AI conversations feel more natural, expressive, and human-like, giving users more control over tone and style. The update could arrive as early as this month.

MAKING GAINS

👩💼 Microsoft Says Everyone Will Be a Boss—of AI

Microsoft predicts a major shift: in the near future, every worker will manage AI agents, not just tasks. In its latest report, Microsoft highlights the rise of “frontier firms”—companies built around AI agents that handle everything from logistics to sales reports. Humans will increasingly act as bosses, setting goals and managing AI “employees” who do the work.

DON’T PANIC

🤖 AI Agents Accumulate Errors in Complex Tasks

While AI agents show promise in automating complex tasks such as data collection, sales outreach, and engineering processes, their reliability remains a concern. Studies have found that these agents often produce errors and hallucinations that worsen with task complexity. For instance, a 1% error rate per step can lead to a 63% probability of failure over 100 steps. Real-world error rates are even higher, potentially around 20% per action.

PLOT TWIST

🎙️ AI Radio Host Fools Listeners for Six Months

For half a year, Australian radio station CADA aired a show hosted by "Thy," an AI-generated voice, without informing listeners. Created using ElevenLabs' technology, Thy presented four hours of music daily, amassing over 72,000 listeners in the final month. The ruse was uncovered when inconsistencies in Thy's speech patterns raised suspicions.