Skip to main content

Apple to Train AI Models Using Blurred Maps Imagery

Apple has revealed plans to leverage imagery from its "Look Around" feature in Apple Maps to train artificial intelligence models. The company updated its official website with details about the initiative, which will begin in March 2025. The move aims to enhance AI capabilities in image recognition, creation, and enhancement across Apple's products and services.

Image

Privacy-First Data Collection

Apple collects visual data through camera-equipped vehicles and backpack-mounted systems for pedestrian areas. All published "Look Around" images already undergo automatic blurring of faces and license plates to protect privacy. The company confirmed it will only use these pre-blurred images for AI training purposes.

Users can additionally request blurring of their residential properties, though this isn't applied by default. "We remain committed to protecting user privacy while improving our services," Apple stated in its announcement.

Existing AI Applications

Several current Apple features already utilize AI-powered image processing:

  • Image Playground: Generative AI for creative visuals
  • Photos Cleanup Tool: Removes unwanted objects from images
  • Advanced Search: Enhanced recognition in the Photos app

The Verge has contacted Apple for specifics about which models will be trained with the Maps imagery and will update their reporting accordingly.

Key Points

  1. Apple will use blurred "Look Around" imagery for AI training starting March 2025
  2. All training data maintains privacy protections with automatic blurring of faces/license plates
  3. Users can request additional blurring of private property images
  4. Multiple existing Apple features already rely on AI image processing technologies

Enjoyed this article?

Subscribe to our newsletter for the latest AI news, product reviews, and project recommendations delivered to your inbox weekly.

Weekly digestFree foreverUnsubscribe anytime

Related Articles

DuckDuckGo Launches Privacy-First AI Voice Chat That Doesn't Store Your Conversations
News

DuckDuckGo Launches Privacy-First AI Voice Chat That Doesn't Store Your Conversations

DuckDuckGo has rolled out a new voice chat feature for its Duck.ai platform that puts privacy front and center. Unlike other voice assistants, this one promises not to store your audio or use it for AI training. Users can chat freely through encrypted channels without creating an account, with OpenAI providing the brains behind the scenes while being contractually barred from keeping any data.

February 11, 2026
PrivacyTechAIEthicsVoiceAssistant
DeepSeek's New OCR Tech Mimics Human Vision, Slashes Costs
News

DeepSeek's New OCR Tech Mimics Human Vision, Slashes Costs

Chinese AI firm DeepSeek has unveiled OCR2, a breakthrough visual encoder that processes documents like human eyes scan pages. By ditching rigid grid processing for flexible 'causal flow tokens,' the system cuts visual token usage by 80% while outperforming Gemini3Pro in benchmarks. The open-sourced technology could pave the way for truly unified multimodal AI.

February 2, 2026
ComputerVisionAIBreakthroughsDocumentAI
OpenClaw: The Lobster AI That Finally Found Its Name
News

OpenClaw: The Lobster AI That Finally Found Its Name

The open-source AI assistant formerly known as Clawd has undergone its third rebranding, settling on OpenClaw after trademark hurdles and community feedback. Despite the naming drama, the project has exploded in popularity, surpassing 100,000 GitHub stars while maintaining its quirky lobster mascot. Offering local AI processing across multiple platforms, OpenClaw lets users manage emails, calendars and more while keeping all data private.

January 30, 2026
AIOpenSourcePrivacyTech
Google's Gemini 3 Flash Now Sees Like a Human Detective
News

Google's Gemini 3 Flash Now Sees Like a Human Detective

Google has upgraded its Gemini 3 Flash AI with groundbreaking 'Agentic Vision' technology that transforms how machines analyze images. Instead of just glancing at pictures, the AI now actively investigates them - zooming in on details, annotating elements, and reasoning like human experts. This breakthrough improves accuracy by 5-10% on complex visual tasks and will soon be available to everyday users through mobile assistants.

January 28, 2026
ComputerVisionGoogleAIImageAnalysis
Robots Can Now Grasp Glassware Thanks to Breakthrough Depth Perception Tech
News

Robots Can Now Grasp Glassware Thanks to Breakthrough Depth Perception Tech

Ant Group's Lingbo Technology has open-sourced LingBot-Depth, a revolutionary spatial perception model that helps robots handle transparent and reflective objects with unprecedented accuracy. Using advanced 'Masked Depth Modeling' technology, the system fills in missing depth data from stereo cameras, solving a longstanding challenge in robotics. Early tests show it outperforms existing solutions by up to 70% in accuracy.

January 27, 2026
RoboticsComputerVisionOpenSource
Kimi K2.5 Sneaks In with Major Visual and Tool Upgrades
News

Kimi K2.5 Sneaks In with Major Visual and Tool Upgrades

Moonshot AI has quietly rolled out Kimi K2.5, bringing significant improvements in visual analysis and tool integration. Users report impressive performance in tasks like converting images to 3D models and solving complex problems step-by-step. The tech community is buzzing with excitement, especially about potential open-source possibilities.

January 27, 2026
AIupdatesComputerVisionMoonshotAI