
AirTag reveals how Amazon destroys rare books for AI training
Amazon buys printed books in bulk, scans them for AI training data, and destroys them in the process.
AINOVAT
Source
60 articles from this source

Amazon buys printed books in bulk, scans them for AI training data, and destroys them in the process.

An Australian user just wanted a spot in a class.

Alibaba's Qwen3.8 Max scores 56 on the Artificial Analysis Intelligence Index, a 10-point jump over Qwen3.7 Max (46).

Anthropic is locking in $10 billion worth of computing capacity from Volta Infra Holdings, a cloud startup that's only a few months old.

One paragraph of "Lord of the Rings" in, 5,500 lines of code out.

Linus Torvalds supports AI tools in Linux kernel development, criticizing those opposed to their use.

Kimi's K3 model has 2.8 trillion parameters, one million tokens of context, and rivals Claude Fable 5 and GPT 5.6 Sol.

Google renames NotebookLM to Gemini Notebook, integrating it deeper into its ecosystem with new coding features and third-party search app connections.

German regulators rule AI Overviews as Google's own content, not neutral search results, crowding out regular links.

OpenAI and Work Louder introduce Codex Micro, a compact controller for managing AI agents with joysticks and a rotary dial.

Sakana AI combines Nvidia's Nemotron models with its Fugu orchestrator to dynamically merge language models for specific tasks.

OpenAI's GPT-Red model finds successful attacks in 84% of test scenarios through self-play training, surpassing human performance.

University of Pennsylvania professor uses GPT-5.6 Sol Pro to disprove Benjamini-Hochberg method conjecture in 90 minutes.

Spotify's AI voice interface now allows Premium subscribers to chat with the service directly inside the app.

OpenAI's Codex tool now encrypts instructions between AI agents, limiting developer visibility into internal task delegation.

Anthropic offers Claude for Teachers, a free AI tool for verified US K-12 educators, with features for lesson planning and student data analysis.

Richard Sutton, 2024 Turing Award winner, launches Oak Lab to build AI agents that learn continuously from their environment.

Google Research's SensorFM uses over a trillion minutes of wearable data to create a general-purpose health intelligence layer.

Anthropic keeps Claude Fable 5 in subscription plans through July 19, 2026, amid pricing pressure from OpenAI's GPT-5.6 Sol.

One in four longer social media posts is entirely AI-generated, according to a Pangram analysis.

S&P Global cuts Oracle's credit rating to 'BBB-', citing OpenAI as a key credit risk.

Meta pulls Muse Image feature that generated AI photos of Instagram users without consent.

Anthropic analyzed 1.2 million Claude Cowork sessions from over 600,000 organizations.

OpenAI CEO Sam Altman now says he's 'pretty sure' AI has created more jobs than it's eliminated.

Brown University professor suspects students used AI to cheat on take-home exam, grades plummet when forced to take in-person test.

Researchers replace growing chat logs with structured memory, enabling AI agents to win at Slay the Spire 2

OpenAI's GPT-5.6 Sol Ultra produces proof of Cycle Double Cover Conjecture in under an hour using 64 subagents.

Boko Haram uses AI chatbots like ChatGPT, Claude, and Gemini to plan attacks, build explosives, and maintain weapons, a Cambridge study finds.

Beijing Academy of Artificial Intelligence releases Orca, a world model predicting abstract world states, trained on 125,000 hours of video.

Meta's Muse Spark 1.1 scored 51 on the Artificial Analysis Intelligence Index, up eight points in three months.

OpenAI acknowledges significant issues with ChatGPT Work and GPT-5.6 Sol, including excessive compute usage and data deletion.

OpenAI kills its AI browser Atlas after eight months, merging features into ChatGPT's Chrome extension.

xAI releases Grok 4.5, trained on tens of thousands of Nvidia GB300 GPUs, with a significantly lower price point.

Mistral introduces Robostral Navigate, an 8B model guiding robots through unknown environments with a single RGB camera.

Anthropic's Claude Fable 5 tops industry benchmarks, but its high pricing may limit adoption.

Microsoft replaces OpenAI and Anthropic AI models with its own MAI models in products like Excel and Outlook.

Cohere releases Transcribe Arabic, an open-source model outperforming Whisper and OmniASR on Arabic dialects and bilingual speech.

Anthropic's AI agent Claude Cowork is now available on mobile and web, expanding beyond its initial desktop app limitation.

China eyes restricting foreign access to its most powerful AI models, affecting Alibaba, Bytedance, and Z.ai.

Anthropic's new tool, J-Lens, can read Claude's internal working memory, revealing its thought process before generating text.

Tencent releases Hy3, an open-source language model with 295 billion parameters built on a mixture-of-experts architecture.

GPT-4 led the Epoch Capabilities Index for about a year, but today's top models barely last seven weeks at the top.

Nvidia's Kyber NVL144 AI server rack delayed to 2028 due to circuit board manufacturing problems.

AWS is ending Mechanical Turk to new customers by July 30, 2026, moving existing users to maintenance mode.

A Google Deepmind developer uses Anthropic's Claude Code to port 2003 game Command & Conquer to native iOS in just a few hours.

Baidu's Unlimited OCR reads dozens of document pages in a single pass, surpassing previous system limits of about ten pages.

Mistral founder Arthur Mensch warns companies against relying on closed AI models.

Bytedance's AI video tool Seedance divides Hollywood.

AI search agents rarely fail at multi-step research due to search, but at asking for clarification on ambiguous queries.

Pxpipe converts long text prompts into compact PNGs, cutting costs by up to 70% for Claude Code and Fable 5 users.

Anthropic developer Thariq Shihipar says Fable 5's bottleneck is the user's blind spots, not the model itself.

Greg Brockman sees a future with invisible, context-aware agents replacing software interfaces.

AI users finished homework faster and scored higher but performed up to 24% worse on exams in a 26,000-student study.

Microsoft plans to merge consumer and enterprise Copilot apps into one in August, cut underused features, and add AI-powered AutoPilot agents.

Anthropic blocks Chinese firms from accessing Claude Code, but they bypass restrictions via VPNs and overseas subsidiaries.

Epoch AI reports a sharp rise in security vulnerability reports, with 1,500 high-severity and critical CVEs in June 2026.

UK's AI Security Institute study reveals standard AI evaluations underestimate agent capabilities by limiting compute budget.

Kuaishou's Kling raises $2 billion from investors, valuing the AI video division at $18 billion ahead of a planned Hong Kong IPO.

Anthropic explores custom AI chip manufacturing with Samsung Electronics.

Anthropic cuts Claude Code's system prompt by 80 percent, citing Fable 5 models' ability to work with fewer instructions.