GKRootWire
Security ICE Signs $2M Deal for Zero-Click Phone Hacking ToolSecurity Attackers Exploit Critical Elementor Pro Bug to Hijack WordPress SitesAI ChatGPT Goes Down, Serves 404 Errors to UsersAI ChatGPT and Codex Suffer Widespread OutageAI Google DeepMind's WeatherNext 3 Sharpens AI Weather ForecastingAI Google's New AI Weather Model Sharpens Storm ForecastsSecurity ICE Signs $2M Deal for Zero-Click Phone Hacking ToolSecurity Attackers Exploit Critical Elementor Pro Bug to Hijack WordPress SitesAI ChatGPT Goes Down, Serves 404 Errors to UsersAI ChatGPT and Codex Suffer Widespread OutageAI Google DeepMind's WeatherNext 3 Sharpens AI Weather ForecastingAI Google's New AI Weather Model Sharpens Storm Forecasts
Gadgets

Xiaomi 17 Ultra's 'Moon Mode' Got Fooled by a Solar Eclipse

Photographers caught Xiaomi's flagship phone applying its AI moon-enhancement trick to a photo of the sun, exposing how much of its 'zoom' photography is synthetic.

During a recent solar eclipse, photographers tested the Xiaomi 17 Ultra's telephoto camera and noticed something odd: the phone's software applied its moon-detail-enhancement algorithm to the sun, producing an image with lunar-like texture instead of an accurate solar image.

Xiaomi, like Samsung before it, uses AI models trained to recognize the moon and overlay or enhance detail that the camera sensor alone can't capture at long zoom ranges. This works fine for actual moon shots, but the system apparently misidentified the sun during the eclipse and applied the same enhancement, revealing that the 'super zoom' moon shots many phones produce are partly AI reconstruction rather than pure optical capture.

The incident reignited debate over how transparent phone makers are about computational photography, especially marketing claims of authentic optical zoom.

Why it matters: This is a concrete, easily reproducible example of AI image enhancement blurring the line between photography and generative reconstruction, which matters as regulators and consumers increasingly question what phone cameras actually capture versus synthesize. It also puts pressure on manufacturers to disclose when scene-recognition AI is substituting trained patterns for real optical detail.

Sources: Hacker News