# AI News Digest - 2026-03-27

1. Google rolled out Search Live globally, deploying a real-time AI search feature that allowed users to interact with Search via voice and camera in more than 200 countries.

2. GitHub announced that it would use Copilot interaction data from Free, Pro, and Pro+ plans to train AI models starting April 24, 2026, unless users actively opted out.

3. Apple obtained full access to Google's Gemini and used distillation techniques to build lightweight on-device AI models for Siri and other device features.

4. Mistral released Voxtral, its first open-weight text-to-speech model, which supported nine languages and could clone voices from approximately three seconds of audio.

# References

1. [https://the-decoder.com/google-rolls-out-search-live-globally-turning-your-phone-camera-into-a-real-time-ai-search-tool/](https://the-decoder.com/google-rolls-out-search-live-globally-turning-your-phone-camera-into-a-real-time-ai-search-tool/)

[Google rolls out Search Live globally, turning your phone camera into a real-time AI search tool](https://the-decoder.com/google-rolls-out-search-live-globally-turning-your-phone-camera-into-a-real-time-ai-search-tool/)

1. [https://the-decoder.com/github-will-use-copilot-interaction-data-to-train-ai-models-starting-april-2026/](https://the-decoder.com/github-will-use-copilot-interaction-data-to-train-ai-models-starting-april-2026/)

[GitHub will use Copilot interaction data to train AI models starting April 2026](https://the-decoder.com/github-will-use-copilot-interaction-data-to-train-ai-models-starting-april-2026/)

1. [https://the-decoder.com/apple-gets-full-gemini-access-and-uses-distillation-to-build-lightweight-on-device-ai/](https://the-decoder.com/apple-gets-full-gemini-access-and-uses-distillation-to-build-lightweight-on-device-ai/)

[Apple gets full Gemini access and uses distillation to build lightweight on-device AI](https://the-decoder.com/apple-gets-full-gemini-access-and-uses-distillation-to-build-lightweight-on-device-ai/)

1. [https://the-decoder.com/mistrals-first-open-weight-tts-model-voxtral-clones-voices-from-three-seconds-of-audio-across-nine-languages/](https://the-decoder.com/mistrals-first-open-weight-tts-model-voxtral-clones-voices-from-three-seconds-of-audio-across-nine-languages/)

[Mistral's first open-weight TTS model Voxtral clones voices from three seconds of audio across nine languages](https://the-decoder.com/mistrals-first-open-weight-tts-model-voxtral-clones-voices-from-three-seconds-of-audio-across-nine-languages/)

For the site tree, see the [root Markdown](https://slashpage.com/ixtj-dev.md).