← archive · dailyblip

Friday, July 17 — six things that changed today

Issue 235 · 2026-07-17

The brief

1. Google Vids now offers custom AI avatars from selfies plus Gemini Omni multi-modal video editing from text prompts and reference images.
2. NYC landlords must now disclose when AI has altered property listing images — a transparency precedent that could spread to other visual advertising sectors.
3. Roblox's new mobile Build feature generates playable games from a single text prompt, opening game creation to anyone without coding skills.
4. Linus Torvalds will not ban AI coding tools from Linux kernel contributions, effectively endorsing AI-assisted development at the OS level.
5. A new survey finds 54% of enterprises have already experienced AI agent security incidents — check your own agent credentials and permission scopes now.
6. DJI faced creator backlash for using an AI-generated human instead of real talent in its Mic Mini ad, intensifying the debate over AI replacing hired performers.

Tool drop

Review by Eddie AI

AI-powered video review tool that lets editors and clients leave time-stamped feedback directly on footage — cuts the back-and-forth for video creators.

The feed that day

xAI sues user for creating CSAM deepfakes with Grok image generator

Elon Musk's xAI filed lawsuit against alleged misuse of its AI image tool to generate child sexual abuse material, marking a rare instance of an AI company suing a user.

rights · via PetaPixel

EU forces Google to share search data and open AI on Android

Digital Markets Act enforcement orders Google to share search results and loosen AI restrictions on Android, raising privacy concerns.

rights · via Ars Technica AI

NYC mayor mandates landlords disclose AI-generated property images

New regulation requires transparency when AI is used to alter real estate listings; part of broader crackdown on deceptive practices.

rights · via PetaPixel

Linus Torvalds will not ban AI coding tools from Linux kernel

Linux creator dismisses calls for AI tool restrictions, says critics should fork the project; signals openness to AI-assisted development.

industry · via Ars Technica AI

Survey: 54% of enterprises report AI agent security incidents

Study finds most companies lack proper credential isolation and scoped identities for AI agents, with majority reporting breaches or near-misses.

industry · via VentureBeat AI

Google Vids adds personalized AI avatars and Gemini Omni editing

Google rolled out custom digital avatars based on selfies and voice, plus multi-modal video generation from text prompts and reference images, with step-by-step editing in Vids.

tools · via TechCrunch AI

DJI faces backlash for AI avatar in TikTok product ad

Creator-focused company used AI-generated human in Mic Mini advertisement, sparking criticism over not hiring actual talent.

rights · via PetaPixel

Roblox launches AI game-builder in mobile app with text prompts

Roblox's new Build feature generates basic playable games from single text prompts, no coding required, powered by open-source and proprietary models.

tools · via TechCrunch AI

NVIDIA Nemotron 3 Embed ranks #1 on RTEB benchmark for agentic retrieval

NVIDIA's embedding model tops retrieval evaluation, advancing RAG infrastructure for agentic workflows.

tools · via Hugging Face Blog

Google AI Mode gains app integrations with Instacart, Canva, YouTube

Google's conversational search now lets users link and execute tasks across select apps, matching ChatGPT and Claude's agentic capabilities.

tools · via TechCrunch AI

Google renames NotebookLM to Gemini Notebook, deepens Gemini integration

AI note-taking app gets rebranding and tighter cross-product integration while remaining standalone.

tools · via The Verge AI

DoorDash opens command-line API for AI agents to search, build carts, order

dd-cli beta enables developers and AI agents to interact with DoorDash via terminal, exemplifying infrastructure built for agentic workflows.

tools · via TechCrunch AI

Moonshot's Kimi K3 with 2–3 trillion parameters expected to rival Anthropic's Opus

Chinese AI lab's upcoming open-weight model aims to close performance gap with frontier closed-source models, releasing imminently.

industry · via TechCrunch AI

China approves Apple Intelligence with Alibaba Qwen and Baidu integration

Apple's generative AI services gain regulatory sign-off in China via partnerships with Alibaba and Baidu for iOS, macOS, and visionOS.

industry · via TechCrunch AI

1Password launches browser integration allowing Claude to access stored credentials

1Password has integrated with Claude, enabling the AI assistant to complete multi-step tasks like travel booking by accessing stored login credentials with user authorization.

tools · via The Verge AI

EU orders Google to open Android and Search to rival AI assistants

European regulators have mandated that Google grant rival AI assistants and search engines comparable access to Android and Search data under the Digital Markets Act.

rights · via The Verge AI

DharmaOCR outperforms newer models on Brazilian Portuguese text recognition

Domain-specialized OCR model demonstrates superior performance through targeted training, competing against larger general-purpose alternatives.

tools · via Hugging Face Blog

Report examines AI sales to police, from data gathering to decision-making systems

Investigation into the commercial deployment of AI systems across police operations, covering surveillance, analytics, and automated decision-making infrastructure.

industry · via The Verge AI