Tech Trends 📰

Cloudflare launches tool to help website owners monetize AI bot crawler access

Cloudflare has launched a new tool that allows website owners to block AI crawlers or charge them using a “pay per crawl” model.. This move aims to help content creators and publishers monetize access to their work as AI bots increasingly collect data without driving traffic back to original sources.. The initiative is backed by major publishers and platforms, including Condé Nast, Associated Press, Reddit, and Pinterest.. Cloudflare’s data shows Google’s crawl-to-referral ratio has worsened (18:1), while OpenAI’s is drastically worse (1,500:1), indicating AI bots are taking content without returning user traffic.. The tool is part of a broader effort to protect online content creators as traditional ad revenue from search traffic declines due to AI-generated summaries and chatbot responses..

Google rolls out its new Veo 3 video-generation model globally | TechCrunch

Google has globally launched its Veo 3 video-generation model in over 159 countries.. The feature is available only to paying subscribers of Google’s AI Pro plan.. Users are limited to generating up to three videos per day using Veo 3.. Veo 3 allows users to create videos up to eight seconds long using text prompts.. Google is also working on adding image-to-video generation capabilities to its Gemini platform..

Cloudflare launches a marketplace that lets websites charge AI bots for scraping | TechCrunch

Cloudflare has launched an experimental marketplace called Pay per Crawl, allowing websites to charge AI bots for accessing their content.. Site owners can set prices for each crawl, block AI bots entirely, or allow them for free — all from a Cloudflare dashboard.. The goal is to give publishers more control and possibly a new revenue stream in the AI era, where bots often scrape content without sending traffic in return.. Cloudflare reports that AI bots like OpenAI’s crawler take far more content than they refer users back — scraping 1,700 times for every one referral.. Several major publishers like TIME, The Atlantic, and Fortune are supporting Cloudflare’s default setting to block AI bots unless granted access..

Grammarly acquires AI email client Superhuman | TechCrunch

Grammarly has acquired AI-powered email app Superhuman to strengthen its productivity tools.. Superhuman raised over $114 million and was last valued at $825 million. Financial terms of the deal were not disclosed.. Superhuman’s CEO Rahul Vohra and the team will join Grammarly.. Grammarly plans to use Superhuman’s technology to build smart AI agents for email, one of Grammarly’s top user needs.. This follows Grammarly’s 2024 acquisition of Coda and a recent $1 billion non-dilutive investment from General Catalyst..

Cursor launches a web app to manage AI coding agents | TechCrunch

Cursor has launched a web app that lets users manage AI coding agents directly from their browser on desktop or mobile.. Users can assign coding tasks (like writing features or fixing bugs) using natural language and monitor progress or merge completed changes.. This follows earlier launches of Cursor’s background agents and a Slack integration, making it easier to delegate tasks via @Cursor.. Cursor is used by over half of the Fortune 500, has reached $500M in annual recurring revenue, and now offers a $200/month Pro tier.. The web app is available to all paying customers with agent access (excluding free tier users) and is part of Cursor’s plan to expand its usefulness beyond just an IDE..

WhatsApp Rolls Out Document Scanning Feature for Android Beta Users: Here’s What to Expect

WhatsApp is testing a new document scanning feature on Android, currently available only for beta users.. The feature lets users scan documents directly within the app using their phone camera, without needing third-party apps.. Scanned documents are saved as PDFs and can be shared instantly with contacts.. The option appears in the attachment menu as "Scan Document" alongside "Browse Documents" and "Choose from Gallery." It supports both manual and automatic scanning modes.. This update, along with WhatsApp’s recent AI-powered chat summarization tool, aims to make the app more useful for productivity and communication..

AI breakthrough helps detect hidden apps on smartphones | The Express Tribune

Australian researchers have developed a new AI method that can detect hidden "vault" apps on Android phones with up to 98% accuracy.. Vault apps hide files, messages, or even other apps using encryption and often look like normal apps, making them hard to spot.. The new technique uses machine learning and doesn't need a list of known apps to find them, unlike current tools.. This could help police and investigators uncover illegal or suspicious content hidden on smartphones.. Researchers plan to test more AI models, expand their data, and apply the method to non-Android devices in the future..

As AI kills search traffic, Google launches Offerwall to boost publisher revenue | TechCrunch

Google has launched Offerwall, a new feature in Ad Manager that helps publishers earn revenue through options like micropayments, surveys, watching ads, and newsletter sign-ups, as traditional search traffic declines due to AI.. The Offerwall uses AI to decide when to show options to users, though publishers can set their own rules for when it appears.. Offerwall is free to use and was tested for over a year with 1,000 publishers; it integrates directly with Google Ad Manager and allows branding customization.. Google is partnering with third-party provider Supertab to allow visitors to pay for temporary content access, including options for subscriptions and content personalization.. During testing, publishers using Offerwall saw average revenue increases of 5% to 15%, with India’s Sakal Media Group reporting a 20% boost and 2 million extra impressions in just three months..

Google launches Doppl, a new app that lets you visualize how an outfit might look on you

Google has launched a new experimental app called Doppl that allows users to visualize how different outfits would look on them using AI.. Users upload a full-body photo of themselves and can then try on outfits virtually by using images or screenshots of clothing they find online or in-person.. The app creates an AI-generated image of the user wearing the selected outfit and can even generate short videos to show how the clothing might look in motion.. Doppl builds on Google Shopping’s virtual try-on tools but focuses on animating the clothes onto the user’s own body instead of generic models.. Currently available in the U.S. on iOS and Android, Doppl is a Google Labs experiment and may have limitations in accurately showing fit and clothing details..

New data highlights the race to build more empathetic language models | TechCrunch

LAION released EmoNet, an open-source toolset focused on detecting and interpreting emotions from voice and images, reflecting a growing industry shift toward developing emotionally intelligent AI.. Emotional intelligence benchmarks like EQ-Bench show that major AI models (from OpenAI, Google, Anthropic, and others) now outperform humans in psychometric tests, with models averaging over 80% accuracy.. Emotional awareness is increasingly influencing AI model rankings and user preference scores, pushing labs to fine-tune their models to understand and respond to complex human emotions.. While emotionally intelligent AI can enhance user experience and mental health monitoring, it also raises safety concerns, such as emotional manipulation or users developing unhealthy attachments.. Experts argue that emotional intelligence in AI could help prevent manipulative behavior, but it requires thoughtful training to avoid unintended consequences and ensure ethical AI development..

Google unveils Gemini CLI, an open source AI tool for terminals | TechCrunch

Google has launched Gemini CLI, an open-source AI tool that allows developers to interact with Gemini models directly from their terminal for tasks like code explanation, debugging, and feature generation.. Gemini CLI runs locally and integrates with developers’ codebases, offering natural language capabilities and supporting other tasks like video creation, research reports, and real-time search.. It’s open-sourced under the permissive Apache 2.0 license to encourage community contributions and developer adoption on GitHub.. Google provides generous usage limits: 60 model requests per minute and 1,000 per day for free users, about twice the average usage rate.. While the tool enhances developer productivity, studies and surveys show concerns about AI coding tools' reliability, with only 43% of developers fully trusting their accuracy..

Meta is adding AI-powered summaries to WhatsApp | TechCrunch

Meta is introducing an AI-powered message summaries feature in WhatsApp to help users quickly review unread messages.. The summaries are private, visible only to the user, and generated using Meta’s “Private Processing” technology, ensuring end-to-end encryption and user privacy are preserved.. The feature is off by default but can be activated through Settings > Chats > Private Processing, where users can manage individual AI functions.. Initially, the feature is rolling out in the U.S. with English support, with plans to expand to more languages and regions later in the year.. Meta AI uses on-device context to generate summaries or writing suggestions without sending or exposing chat content to Meta or WhatsApp servers..