AI training Archives

TrendTechie

September 2, 2026

Sony и Warner подают в суд на Anthropic из‑за обучения Claude на пиратских песенниках

Sony Music Publishing and Warner Chappell Music have filed a lawsuit against Anthropic, claiming that the company used pirated songbooks and song lyrics from "shadow libraries" to train its AI model, Claude. The lawsuit cites approximately 7 million books obtained through torrent downloads and web scraping, including unauthorized sheet music and songbooks. The plaintiffs are seeking statutory damages of up to 0,000 for each work that is found to have been willfully infringed. The lawsuit was filed on August 28, 2026, in the U.S. District Court for the Northern District of California, naming Anthropic and its co-founders as defendants. The plaintiffs allege that the data sources used for training included 5 million books downloaded from Library Genesis and 2 million from Pirate Library Mirror, as well as scraped song lyrics from licensed platforms Musixmatch and LyricFind. The use of BitTorrent for downloads introduces legal risks due to the simultaneous distribution of files. The lawsuit emphasizes the distinction between legally purchased and pirated materials, with prior cases indicating that downloads from pirate libraries constitute significant copyright infringement.

AppWizard

August 28, 2026

Google gives Android developers until February to use less RAM

AI data centers are consuming memory chips faster than the semiconductor industry can adjust production, affecting the Android ecosystem. Google announced new memory reduction requirements for all applications on Google Play due to hardware supply constraints impacting device memory availability. Every Android application must comply with revised performance thresholds for dynamic memory usage and bitmap allocation, with developers exceeding limits receiving automated notifications. A Memory Limiter tool will be introduced to enforce compliance by restricting non-compliant applications' memory usage. Budget Android devices, particularly those with 4GB of RAM, are at greater risk of crashes without optimization from developers. This policy will impact millions of users in regions with affordable Android phones. Chip foundries like Samsung, SK Hynix, and Micron are reallocating production capacity towards High Bandwidth Memory for AI training chips, leading to a tighter supply of conventional DRAM for consumer devices. Google has not specified numerical thresholds for compliance, and the consequences for non-compliance remain unclear, with only warnings mentioned. The Memory Limiter tools are expected to be rolled out before the end of 2026, with full enforcement starting in February 2027, and a second requirement, Zero Tap Sign-In, will be introduced in April 2027. Development teams should prepare for these changes in the latter half of 2026.

TrendTechie

July 18, 2026

Anthropic снова в суде: компания не только обучала ИИ на пиратских книгах, но и сама скачивала их через торренты

More than 100 authors have filed a lawsuit against Anthropic, claiming over 0 million for the unauthorized use of their books in training AI systems. The complaint was submitted to the federal court for the Northern District of California on June 17, alleging that Anthropic unlawfully obtained and utilized over 500 pirated copies of their works. The lawsuit includes various literary works and notable plaintiffs, such as Nolan Bushnell and Donna Barba Igna, who are demanding 0,000 for each work used without permission. The authors claim Anthropic downloaded books via BitTorrent and used illegal libraries, integrating these works into its AI training systems. This lawsuit follows a previous class action against Anthropic that concluded with a .5 billion settlement. The current plaintiffs have opted out of that settlement and are pursuing individual claims, asserting that Anthropic not only used but also distributed pirated copies. The lawsuits estimate around seven million works may have been unlawfully utilized in AI development.

AppWizard

June 3, 2026

Google Is Quietly Buying Code From Play Store Developers to Train AI

Google has reached out to select Android app developers with a proposal to monetize their codebases through a "confidential content offer pilot." Developers are invited to share their active production code and archived projects, allowing them to generate additional revenue while retaining full intellectual property rights. The licensing agreement is non-exclusive. The initiative aims to enhance Google's AI coding tools by utilizing real-world code from developers, including unused prototypes and side projects. This move comes as Google faces competition in the AI coding space from companies like Anthropic and Microsoft.

AppWizard

May 9, 2026

Instagram Loses End-To-End Message Encryption Today; What It Means For You

Instagram has discontinued its end-to-end encryption (e2ee) feature for direct messages, which previously allowed users to communicate securely without interception. All direct messages will now be protected by standard encryption, allowing potential access by service or network providers. Meta, Instagram's parent company, cited low usage rates for this change, which was communicated in March. Privacy advocates have raised concerns about user communications being shared with law enforcement and for AI training purposes, although Meta clarified it does not use private messages for AI purposes. Users seeking privacy can switch to WhatsApp or the standalone Messenger app, which still support e2ee. Meta has also advised users who had e2ee enabled to download their chat histories and media before the feature is fully retired.

AppWizard

April 29, 2026

How to turn off Meta AI (Facebook, Instagram, and WhatsApp)

Meta AI has been integrated into Facebook, Instagram, WhatsApp, and Messenger, enhancing user experiences with features like search functionalities, chat interactions, and content generation. Users express a desire to limit their interactions with Meta AI due to concerns about transparency, privacy, and data security. The AI model is trained on diverse data sources, including public posts and user interactions, raising issues regarding data control. On WhatsApp, personal messages and calls are end-to-end encrypted, while Instagram may share past messages with Meta AI for context. Security vulnerabilities have been reported, including incidents of data mishandling. Currently, there is no comprehensive option to disable Meta AI across all platforms, but users can take steps to limit interactions, such as muting AI prompts and adjusting privacy settings. Users in the E.U. and U.K. can object to certain data uses under GDPR by submitting requests through their Meta accounts. To protect privacy, users are advised to opt out of AI training, limit personal data sharing, and use privacy tools like VPNs.

Winsage

April 14, 2026

Veteran Windows dev shows off AI running on 47-year-old PDP11 with 6 MHz CPU and 64KB of RAM — ‘gloriously absurd’ project runs transformer model written in PDP-11 assembly language

Veteran Windows developer Dave Plummer is using a 47-year-old PDP-11 system with a 6 MHz CPU and 64KB of RAM to train a neural network called ‘Attention 11,’ developed in PDP-11 assembly language. The model is tasked with reversing a sequence of eight digits, requiring it to understand structural rules similar to those used in contemporary large language models like ChatGPT. Plummer emphasizes the importance of optimizing for the system's constraints, stating that constraints drive creative engineering. The model operates with 1,216 parameters and uses fixed-point math with 8-bit precision. After approximately 350 training steps, the model achieved 100% accuracy in the number-reversing task in about 3.5 minutes. Plummer argues that modern AI operates on the same mechanical principles as this vintage machine, just at a much larger scale, and suggests that companies focusing on efficiency and optimization may gain an advantage in the AI landscape.

AppWizard

February 17, 2026

The Simplest Android App for Scanning Documents

FairScan is a free and open-source scanning application designed for Android users, allowing them to photograph, crop, and compile multipage documents into a single PDF file. The app prioritizes user privacy and simplicity, avoiding intrusive ads and questionable privacy practices common in other scanning apps. Users can download FairScan from the Google Play Store or F-Droid, and the scanning process involves capturing images of documents in a well-lit area, with the option to add additional pages. Scanned documents can be exported as a single PDF or multiple JPEG files. While FairScan lacks features like post-capture editing and optical character recognition, it effectively serves its primary purpose without unnecessary distractions.

AppWizard

February 11, 2026

I simplified my Google Photos searching and editing with this trick

Google Photos has introduced a new feature called Ask Photos, which allows users to edit images using natural-language prompts instead of traditional editing tools. This feature is available for personal Google accounts in the U.S. and requires users to opt in and enable face match and location estimation. While Ask Photos simplifies the editing process, users should be cautious about privacy, as prompts will be used for training purposes, even though Google does not use photos or videos for AI training. To use Ask Photos, users must open the Google Photos app, select a photo, tap the Edit button, and enter their request in the Help me edit text field. Additionally, Google Photos provides transparency regarding edits through C2PA Content Credentials, which document the types of edits made and distinguish between original and edited photos.

AppWizard

January 27, 2026

BETT 2026: Microsoft reveals major Copilot, AI, and Minecraft Education updates | ETIH EdTech News

Microsoft showcased a comprehensive update to its education platform at BETT 2026, integrating AI tools like Copilot into Microsoft 365, Teams for Education, Learning Accelerators, and Minecraft Education. The new Teach module allows educators to create lesson plans aligned with standards from over 35 countries, generate rubrics, and modify reading levels. Minecraft Education lesson plan creation using Copilot is set to enter preview in February 2026. Microsoft introduced a Study and Learn agent for students, designed to support independent learning through tailored interactive activities, with a preview scheduled for January 2026. The program emphasizes AI literacy through Minecraft Education, helping students recognize AI in tools, understand algorithms, and develop critical evaluation skills. Learning Accelerators will expand with a new Microsoft 365 LTI for LMS integration, allowing full integration of Microsoft 365 apps and Copilot into learning management systems by Spring 2026. Microsoft also addressed AI misuse concerns, introducing assignment-level controls for educators to define acceptable AI usage, with previews scheduled for February 2026.