We’re bringing Gemini’s state-of-the-art translation model to Google Translate for text, and more new features.
Quantum computing holds the potential to solve some of the world's most complex challenges. Today, we’re announcing a new collaboration with the UK National Quantum Comp…
Quantum computing holds the potential to solve some of the world's most complex challenges. Today, we’re announcing a new collaboration with the UK National Quantum Comp…
Earlier this year, we reiterated our commitment to keeping Android and Google Play safe for everyone and maintaining a thriving environment where users can trust the apps they download and your business can flourish. We’ve heard your feedback clearly, from excited conversations at Play events around the world to the honest concerns on social media. You want simpler ways to make sure your apps are compliant and pass review, and need strong protections for your business so you can focus on growth and innovation. We are proud of the steps we’ve taken together this year, but know this is ongoing work in a complex, ever-changing market.
Here are key actions we’ve taken this year to simplify your development journey and strengthen protection.
Policy guidance right where you code: We rolled out Play Policy Insights to all developers using Android Studio. This feature provides real-time, in-context guidance and policy warnings as you code, helping you proactively identify and resolve potential issues before you even submit your app for review.
Pre-review checks to help prevent app review surprises: Last year, we launched pre-review checks in Play Console so you can identify issues early, like incomplete policy declarations or crashes, and avoid rejections. This year, we expanded these checks for privacy policy links, login credential requirements, data deletion request links, inaccuracies in your Data safety form, and more.
Advanced abuse and fraud protection: We made the Play Integrity API faster and more resilient, and introduced new features like Play remediation prompts and device recall in beta. Device recall is a powerful new tool that lets you store and recall limited data associated with a device, even if the device is reset, helping protect your business model from repeat bad actors.
Tools to keep kids safe:
We continued to invest in protecting children across Google products, including Google Play. New Play policy helps keep our youngest users safe globally by requiring apps with dating and gambling features to use Play Console tools to prevent minors from accessing them. Our enhanced Restrict Minor Access feature now blocks the users who we determine to be minors from searching for, downloading, or making purchases in apps that they shouldn’t have access to.
We’ve also been providing tools to developers to help meet significant new age verification regulatory requirements in applicable US states.
More ways to stop malware from snooping on your app: Android 16 provides a new, powerful defense in a single line of code: accessibilityDataSensitive. This flag lets you explicitly mark views in your app as containing sensitive data and block malicious apps from seeing or performing interactions on it. If you already use setFilterTouchesWhenObscured(true) to protect your app from tapjacking, your views are automatically treated as sensitive data for accessibility for an instant additional layer of defense with no extra work.
More support when you need it: Beyond the webinars and resources that we share, you told us you needed more direct policy help to understand requirements and get answers. Next week, we’ll add a direct way for you to reach our team about policy questions in your Play Console. You’ll be able to find this new, integrated support experience directly within your Play Console via the “Help” section. We also expanded the Google Play Developer Help Community to more languages, like Indonesian, Japanese, Korean, and Portuguese.
Clearer documentation: You asked for policy that’s easier to understand. To help you quickly grasp essential requirements, we've introduced a new Key Considerations section across several policies (like Permissions and Target API Level) and included concise "Do's & Don'ts" and easier-to-read summaries.
More transparent appeals process: We introduced a 180-day appeal window for account terminations. This allows us to prioritize and make decisions faster for developers who file appeals.
Android developer verification design changes: To support a diverse range of users and developers, we’re taking action on your feedback.
First, we’re creating a dedicated free account type to support students and hobbyists who want to build apps just for a small group, like family and friends. This means that you can share your creations to a limited number of devices without needing to go through the full developer verification process.
We’re also building a flow for experienced users to be able to install unverified apps. This is being carefully designed to balance providing choice with prioritizing security, including clear warnings so users fully understand the risks before choosing to bypass standard safety checks.
The improvements we made this year are only the beginning. Your feedback helps drive our roadmap, and it will continue to inform future refinements to our policies, tools, experiences, and ensuring Android and Google Play remain the safest and most trusted place for you to innovate and grow your business.
Thank you for being our partner in building the future of Android.
Earlier this year, we reiterated our commitment to keeping Android and Google Play safe for everyone and maintaining a thriving environment where users can trust the apps they download and your business can flourish. We’ve heard your feedback clearly, from excited conversations at Play events around the world to the honest concerns on social media. You want simpler ways to make sure your apps are compliant and pass review, and need strong protections for your business so you can focus on growth and innovation. We are proud of the steps we’ve taken together this year, but know this is ongoing work in a complex, ever-changing market.
Here are key actions we’ve taken this year to simplify your development journey and strengthen protection.
Policy guidance right where you code: We rolled out Play Policy Insights to all developers using Android Studio. This feature provides real-time, in-context guidance and policy warnings as you code, helping you proactively identify and resolve potential issues before you even submit your app for review.
Pre-review checks to help prevent app review surprises: Last year, we launched pre-review checks in Play Console so you can identify issues early, like incomplete policy declarations or crashes, and avoid rejections. This year, we expanded these checks for privacy policy links, login credential requirements, data deletion request links, inaccuracies in your Data safety form, and more.
Advanced abuse and fraud protection: We made the Play Integrity API faster and more resilient, and introduced new features like Play remediation prompts and device recall in beta. Device recall is a powerful new tool that lets you store and recall limited data associated with a device, even if the device is reset, helping protect your business model from repeat bad actors.
Tools to keep kids safe:
We continued to invest in protecting children across Google products, including Google Play. New Play policy helps keep our youngest users safe globally by requiring apps with dating and gambling features to use Play Console tools to prevent minors from accessing them. Our enhanced Restrict Minor Access feature now blocks the users who we determine to be minors from searching for, downloading, or making purchases in apps that they shouldn’t have access to.
We’ve also been providing tools to developers to help meet significant new age verification regulatory requirements in applicable US states.
More ways to stop malware from snooping on your app: Android 16 provides a new, powerful defense in a single line of code: accessibilityDataSensitive. This flag lets you explicitly mark views in your app as containing sensitive data and block malicious apps from seeing or performing interactions on it. If you already use setFilterTouchesWhenObscured(true) to protect your app from tapjacking, your views are automatically treated as sensitive data for accessibility for an instant additional layer of defense with no extra work.
More support when you need it: Beyond the webinars and resources that we share, you told us you needed more direct policy help to understand requirements and get answers. Next week, we’ll add a direct way for you to reach our team about policy questions in your Play Console. You’ll be able to find this new, integrated support experience directly within your Play Console via the “Help” section. We also expanded the Google Play Developer Help Community to more languages, like Indonesian, Japanese, Korean, and Portuguese.
Clearer documentation: You asked for policy that’s easier to understand. To help you quickly grasp essential requirements, we've introduced a new Key Considerations section across several policies (like Permissions and Target API Level) and included concise "Do's & Don'ts" and easier-to-read summaries.
More transparent appeals process: We introduced a 180-day appeal window for account terminations. This allows us to prioritize and make decisions faster for developers who file appeals.
Android developer verification design changes: To support a diverse range of users and developers, we’re taking action on your feedback.
First, we’re creating a dedicated free account type to support students and hobbyists who want to build apps just for a small group, like family and friends. This means that you can share your creations to a limited number of devices without needing to go through the full developer verification process.
We’re also building a flow for experienced users to be able to install unverified apps. This is being carefully designed to balance providing choice with prioritizing security, including clear warnings so users fully understand the risks before choosing to bypass standard safety checks.
The improvements we made this year are only the beginning. Your feedback helps drive our roadmap, and it will continue to inform future refinements to our policies, tools, experiences, and ensuring Android and Google Play remain the safest and most trusted place for you to innovate and grow your business.
Thank you for being our partner in building the future of Android.
Gradient Canvas is a new art exhibition celebrating a decade of creative collaborations between artists and artificial intelligence.
The ChromeOS Beta channel is being updated to OS version 16463.45.0 (Browser version 143.0.7499.114) for most ChromeOS devices.
Visit our ChromeOS communities
General: Chromebook Help Community
Beta Specific: ChromeOS Beta Help Community
Interested in switching channels? Find out how.
Luis Menezes
Google ChromeOS
The Dev channel has been updated to 145.0.7572.3 for Windows, Mac and Linux.
A partial list of changes is available in the Git log. Interested in switching release channels? Find out how. If you find a new issue, please let us know by filing a bug. The community help forum is also a great place to reach out for help or learn about common issues.
Chrome Release Team
Google Chrome
In the rapidly evolving world of AI models for mobile devices, a persistent challenge is how to bring SOTA LLMs to smartphones without compromising on privacy or requiring App developers to be Machine Learning engineers.
Today, we are excited to talk about how Cactus, a startup building a next-gen inference engine for mobile devices, fine-tunes the open-source Gemma 3 model. By leveraging Tunix, the LLM post-training library in the JAX ML ecosystem, they achieved this entirely on Google Colab's Free Tier.
For app developers, running Large Language Models (LLMs) in the cloud isn't always an option due to privacy concerns (like GDPR) and latency requirements. The solution lies in running models locally on the device. However, most smartphones globally lack specialized MPUs (Micro Processing Units), meaning developers need highly efficient, smaller models.
While compact models like Gemma (270M or 1B parameters) are incredibly efficient, they are often "generalists." To be useful for specific mobile applications—such as a medical imaging assistant or a legal document analyzer—they need to be fine-tuned to become domain experts.
The problem? Most app developers are not ML infrastructure experts. Setting up complex training pipelines, managing dependencies, and navigating steep learning curves creates too much friction.
To solve this, Cactus created a simplified "Low-Friction" workflow by implementing a Python script using Supervised Fine Tuning (SFT) APIs of Tunix in a Colab.
1. The Engine: Tunix
Cactus utilized Tunix, Google's lightweight and modular LLM post-training library, which supports both SFT and leading RL algorithms, and executes natively on TPUs. Tunix strips away the complexity of heavy frameworks, offering a simplified path to Supervised Fine-Tuning (SFT).
2. The Access: Google Colab Free Tier
Accessibility was a key requirement. Instead of requiring developers to set up complex cloud billing and project IDs immediately, the workflow operates entirely within a Google Colab Notebook. By utilizing the free tier of Colab, developers can:
3. The Deployment: Cactus
Once tuned, the model is converted into the Cactus graph format. This allows the now-specialized Gemma 3 model to be deployed directly into a Flutter or native mobile app with just a few lines of code, running efficiently on a wide range of smartphone hardware.
"Our users are app developers, not ML engineers," explains Henry Ndubuaku, co-founder of Cactus. "They want to pick a model, upload data, and click 'tune.' By using Tunix and Colab, we can give them a 'clone-and-run' experience that removes the intimidation factor from fine-tuning."
This workflow represents the "lowest hanging fruit" in democratizing AI:
While the Colab notebook provides an immediate, accessible solution, Cactus is exploring a future plan to build a full GUI-based portal for fine-tuning and quantization of LLMs with the back end compute as Google Cloud TPUs, allowing for scalable training of larger models and even more seamless integration into the mobile development lifecycle.
Get Started
Ready to turn your mobile app into an AI powerhouse? Check out the Tunix SFT Notebook for Cactus and start fine-tuning Gemma 3 for your device today:
You can explore Tunix sample scripts, documentation and repo at:
In the rapidly evolving world of AI models for mobile devices, a persistent challenge is how to bring SOTA LLMs to smartphones without compromising on privacy or requiring App developers to be Machine Learning engineers.
Today, we are excited to talk about how Cactus, a startup building a next-gen inference engine for mobile devices, fine-tunes the open-source Gemma 3 model. By leveraging Tunix, the LLM post-training library in the JAX ML ecosystem, they achieved this entirely on Google Colab's Free Tier.
For app developers, running Large Language Models (LLMs) in the cloud isn't always an option due to privacy concerns (like GDPR) and latency requirements. The solution lies in running models locally on the device. However, most smartphones globally lack specialized MPUs (Micro Processing Units), meaning developers need highly efficient, smaller models.
While compact models like Gemma (270M or 1B parameters) are incredibly efficient, they are often "generalists." To be useful for specific mobile applications—such as a medical imaging assistant or a legal document analyzer—they need to be fine-tuned to become domain experts.
The problem? Most app developers are not ML infrastructure experts. Setting up complex training pipelines, managing dependencies, and navigating steep learning curves creates too much friction.
To solve this, Cactus created a simplified "Low-Friction" workflow by implementing a Python script using Supervised Fine Tuning (SFT) APIs of Tunix in a Colab.
1. The Engine: Tunix
Cactus utilized Tunix, Google's lightweight and modular LLM post-training library, which supports both SFT and leading RL algorithms, and executes natively on TPUs. Tunix strips away the complexity of heavy frameworks, offering a simplified path to Supervised Fine-Tuning (SFT).
2. The Access: Google Colab Free Tier
Accessibility was a key requirement. Instead of requiring developers to set up complex cloud billing and project IDs immediately, the workflow operates entirely within a Google Colab Notebook. By utilizing the free tier of Colab, developers can:
3. The Deployment: Cactus
Once tuned, the model is converted into the Cactus graph format. This allows the now-specialized Gemma 3 model to be deployed directly into a Flutter or native mobile app with just a few lines of code, running efficiently on a wide range of smartphone hardware.
"Our users are app developers, not ML engineers," explains Henry Ndubuaku, co-founder of Cactus. "They want to pick a model, upload data, and click 'tune.' By using Tunix and Colab, we can give them a 'clone-and-run' experience that removes the intimidation factor from fine-tuning."
This workflow represents the "lowest hanging fruit" in democratizing AI:
While the Colab notebook provides an immediate, accessible solution, Cactus is exploring a future plan to build a full GUI-based portal for fine-tuning and quantization of LLMs with the back end compute as Google Cloud TPUs, allowing for scalable training of larger models and even more seamless integration into the mobile development lifecycle.
Get Started
Ready to turn your mobile app into an AI powerhouse? Check out the Tunix SFT Notebook for Cactus and start fine-tuning Gemma 3 for your device today:
You can explore Tunix sample scripts, documentation and repo at:
