AI Daily Briefing — 2026-05-13

Google has announced a new Android feature called 'Create My Widget,' which allows users to build custom widgets using natural language commands. The tool, set to debut this summer on select Samsung Galaxy and Google Pixel devices, lets people describe exactly what they want—such as a weekly meal prep dashboard or a weather display showing only wind and rain—and Gemini will generate a personalized, resizable widget for the home screen.

Gemini can pull data from the web and integrate with Google apps like Gmail and Calendar. For example, a user planning a family reunion in Berlin could get a single widget that shows flight info, hotel bookings, restaurant reservations, and a countdown timer. Google says this is part of a broader effort to make generative AI a core part of the Android experience, enabling everyday users to create tools that feel like a personal assistant working on repeat.

Ben Greenwood, director of product management for Android Core Experiences, explained that the feature combines Gemini's knowledge of the world with personal data to unlock countless use cases. The announcement came alongside the unveiling of Gemini Intelligence, which will bring AI-powered autofill and voice dictation to Gboard. The move signals Google's push to make customization more accessible as tech companies race to embed AI into mobile interfaces.

Months before winning the 2024 Nobel Prize in economics, Daron Acemoglu published a paper that challenged Silicon Valley’s optimism. He argued artificial intelligence would only modestly boost U.S. productivity and wouldn’t replace human workers. Two years later, despite rapid AI advances, the data largely supports his cautious view. In an interview with MIT Technology Review AI, Acemoglu outlined three key trends he’s watching now.

Meanwhile, tech icon Stewart Brand is championing a different kind of revolution: maintenance. In his new book, “Maintenance: Of Everything, Part One,” Brand frames upkeep as a civilizational act. He argues that caring for objects, infrastructure, or the planet can be radical. Yet critics note his vision feels more personal than collective, missing the chance to improve shared spaces.

Separately, Google detected the first known zero-day exploit created entirely by AI. The attack was stopped before causing damage, but experts warn AI-powered hacking is becoming an industrial-scale threat. OpenAI also launched a new tool to rival Anthropic’s Claude Mythos, aiming to patch vulnerabilities faster than attackers can find them.

In other news, Ilya Sutskever testified in the Altman v. Musk trial, describing a “pattern of lying” by Sam Altman. Texas sued Netflix over alleged data harvesting and addictive design. And Moderna is developing an mRNA vaccine for hantavirus, following a cruise ship outbreak.

A recent hantavirus outbreak on a Dutch-flagged cruise ship has resulted in eight passengers contracting the Andes virus, with three fatalities. Health experts emphasize this is not comparable to the COVID-19 pandemic, as transmission requires close contact, which the ship's environment enabled. No specific treatments or vaccines exist, but containment is considered feasible due to the virus's limited spread patterns, according to MIT Technology Review AI.

In the second week of the Elon Musk versus OpenAI trial, testimony revealed Musk previously pushed for a for-profit structure at OpenAI. Shivon Zilis testified that Musk attempted to recruit Sam Altman for a new AI venture. Court proceedings also included details of Greg Brockman's private journals and Musk's abandoned plans for a rival AI lab, highlighting the intense legal battle over the company's direction.

Separately, researchers warn that large language models could enhance mass surveillance by quickly linking anonymized data to real individuals. Privacy experts fear this capability removes existing barriers protecting the public, as data brokers already sell personal information to government clients. The technology's scalability raises concerns about erosion of privacy protections.

Other notable tech stories include Meta employees feeling pressured to adopt AI amid layoff fears, South Korea's military exploring robots to address troop shortages, and ongoing lawsuits against OpenAI over ChatGPT's alleged role in harmful content generation.

Google announced a suite of new AI features for Android at its I/O Edition event on Tuesday, expanding the capabilities of its Gemini Intelligence assistant. The updates enable the AI to perform complex, multi-step tasks across apps, such as copying a grocery list from a notes app and adding items to a shopping cart. Users can activate these actions by pressing the power button and describing the task, with the phone’s screen providing context. Google emphasized that Gemini will require final confirmation before completing any checkout process.

Additional features include an auto-browse function that lets Gemini navigate the web to book appointments or fill out forms, using personal details learned through an opt-in Personal Intelligence system. Later this year, Android devices will gain Gemini in Chrome for summarizing web pages, and the Gboard keyboard will introduce Rambler, a dictation tool that transcribes speech and removes filler words while preserving the user’s tone.

Google is also bringing vibe-coding to Android, allowing users to create custom widgets by describing them in natural language. For example, a user could generate a meal-planning widget with a query like “Suggest three high-protein meal prep recipes every week.” These widgets will adhere to Google’s Material 3 design language. The AI-powered features will first roll out to Samsung Galaxy and Google Pixel devices this summer, with broader Android availability later in the year.

Google unveiled Rambler, an AI-powered voice dictation tool for its Gboard keyboard, during the Android Show: I/O Edition 2026 event. The feature, built on Gemini models, directly challenges emerging dictation startups like Wispr Flow and Typeless by offering a native, pre-installed alternative on Android devices. Rambler cleans up speech by removing filler words and understanding mid-sentence corrections, such as changing a meeting time from 3 p.m. to 2 p.m. on the fly.

A standout capability is code-switching, which allows users to seamlessly alternate between languages like English and Hindi without losing context. This multilingual support, powered by Gemini, addresses a gap many Western dictation apps have overlooked. Google emphasized privacy, stating Rambler does not store voice recordings and uses a mix of on-device and cloud processing. The company aims to position the feature as a keyboard reinvention, available across all apps.

Initially rolling out on Samsung Galaxy and Google Pixel phones this summer, Rambler will eventually expand to other Android devices. For startups, the challenge is steep: Gboard is the default keyboard for hundreds of millions of users, giving Google unmatched distribution. Independent dictation apps now must prove superior accuracy, deeper features, or stronger privacy to justify a separate download. The question is no longer about building a good product, but whether it can be compelling enough to seek out.

Source: TechCrunch AI