Google

Gemini 2.5 Computer Use brings human-like interface control to AI agents

Google DeepMind has launched the Gemini 2.5 Computer Use model, a specialised version of Gemini 2.5 Pro designed to let AI agents interact directly with digital user interfaces.

Available in preview through the Gemini API, developers can build agents capable of performing web and mobile tasks such as form-filling, navigation and interaction within apps.

Unlike models limited to structured APIs, Gemini 2.5 Computer Use can reason visually about what it sees on screen, making it possible to complete tasks requiring clicks, scrolls and text input.

While maintaining low latency, it outperforms rivals on several benchmarks, including Browserbase’s Online-Mind2Web and WebVoyager.

The model’s safety design includes per-step risk checks, built-in safeguards against misuse and developer-controlled restrictions on high-risk actions such as payments or security changes.

Google has already integrated it into systems like Project Mariner, Firebase Testing Agent and AI Mode in Search, while early testers report faster, more reliable automation.

Gemini 2.5 Computer Use is now available in public preview via Google AI Studio and Vertex AI, enabling developers to experiment with advanced interface-aware agents that can perform complex digital workflows securely and efficiently.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Google unveils CodeMender, an AI agent that repairs code vulnerabilities

Google researchers have unveiled CodeMender, an AI-powered agent designed to automatically detect and fix software vulnerabilities.

The tool aims to improve code security by generating and applying patches that address critical flaws, allowing developers to focus on building reliable software instead of manually locating and repairing weaknesses.

Built on the Gemini Deep Think models, CodeMender operates autonomously, identifying vulnerabilities, reasoning about the underlying code, and validating patches to ensure they are correct and do not introduce regressions.

Over the past six months, it has contributed 72 security fixes to open source projects, including those with millions of lines of code.

The system combines advanced program analysis with multi-agent collaboration to strengthen its decision-making. It employs techniques such as static and dynamic analysis, fuzzing and differential testing to trace the root causes of vulnerabilities.

Each proposed fix undergoes rigorous validation before being reviewed by human developers to guarantee quality and compliance with coding standards.

According to Google, CodeMender’s dual approach (reactively patching new flaws and proactively rewriting code to eliminate entire vulnerability classes) represents a major step forward in AI-driven cybersecurity.

The company says the tool’s success demonstrates how AI can transform the maintenance and protection of modern software systems.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Former Google CEO backs Antarctic drone venture

A reported investment by former Google CEO Eric Schmidt aims to deploy advanced drone systems to navigate Antarctic waters under extreme conditions. The project involves autonomous aerial and underwater drones tailored for polar environments.

Schmidt’s initiative would target the Southern Ocean’s carbon cycle, ice dynamics, and climate modelling. Designers intend drones to operate where traditional vessels cannot, gathering otherwise unreachable data to refine climate models.

Technologies under development reportedly include cold-resistant batteries, autonomous navigation systems, satellite or acoustic communications, and ice-penetrating radar for subsurface mapping. The designs emphasise minimal human intervention.

There is room for application beyond research, including maritime logistics in polar routes and environmental monitoring. If real, the investment could reshape the future of work on how scientists and explorers gather data in remote, hostile regions.

On the other hand, there are criticisms to exploring the area with technologies that could disturb the ecosystem and native species already under other threats. Therefore, careful consideration will have to be made of the ecological impact of this initiative.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot! 

Atlantic Quantum joins Google Quantum AI to advance scalable quantum hardware

Google Quantum AI has taken a major step in its pursuit of error-corrected quantum computing by integrating Atlantic Quantum, an MIT spin-out focused on superconducting hardware.

The move, while not formally labelled an acquisition, effectively brings the startup’s technology and talent into Google’s programme, strengthening its roadmap toward scalable quantum systems.

Atlantic Quantum, founded in 2021, has worked on integrating qubits with superconducting control electronics in the same cold stage.

A modular chip stack that promises to simplify design, reduce noise, and make scaling more efficient. Everything is equally important to build machines capable of solving problems beyond the reach of classical computers.

Google’s Hartmut Neven highlighted the approach as a way to accelerate progress toward large, fault-tolerant devices.

The startup’s journey, from MIT research labs to Google integration, has been rapid and marked by what CEO Bharath Kannan called ‘managed chaos’.

The founding team and investors were credited with pushing superconducting design forward despite the immense challenges of commercialising such cutting-edge technology.

Beyond hardware, Google gains a strong pool of engineers and researchers, enhancing its competitive edge in a field where rivals include IBM and several well-funded scale-ups.

A move that reflects a broader industry trend where research-heavy startups are increasingly folded into major technology firms to advance long-term quantum ambitions. With governments and corporations pouring resources into the race, consolidation is becoming common.

For Atlantic Quantum, joining Google ensures both technological momentum and access to resources needed for the next phase. As co-founder Simon Gustavsson put it, the work ‘does not stop here’ but continues within Google Quantum AI’s effort to deliver real-world quantum applications.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Oracle systems targeted in unverified data theft claims, Google warns

Google has warned that hackers are emailing company executives, claiming to have stolen sensitive data from Oracle business applications. The group behind the campaign identifies itself as affiliated with the Cl0p ransomware gang.

In a statement, Google said the attackers target executives at multiple organisations with extortion emails linked to Oracle’s E-Business Suite. The company stated that it lacks sufficient evidence to verify the claims or confirm whether any data has been taken.

Neither Cl0p nor Oracle responded to requests for comment. Google did not provide additional information about the scale or specific campaign targets.

The cl0p ransomware gang has been involved in several high-profile extortion cases, often using claims of data theft to pressure organisations into paying ransoms, even when breaches remain unverified.

Google advised recipients to treat such messages cautiously and report any suspicious emails to security teams while investigations continue.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!

Bye Bye Google AI hides unwanted AI results in Search

Google is pushing AI deeper into its services, with AI Overviews already reaching billions of users and AI Mode now added to Search. Chrome is also being rebranded as an AI-first browser.

Not all users welcome these changes. Concerns remain about accuracy, intrusive design and Google’s growing control over how information is displayed. Unlike other features, AI elements in Search cannot be turned off directly, leaving users reliant on third-party solutions.

One such solution is the new ‘Bye Bye, Google AI’ extension, which hides AI-generated results and unwanted blocks such as sponsored links, shopping sections and discussion forums.

The extension works across Chromium-based browsers, though it relies on CSS and may break when Google updates its interface.

A debate that reflects wider unease about AI in Search.

While Google claims it improves user experience, critics argue it risks spreading false information and keeping traffic within Google’s ecosystem rather than directing users to original publishers.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Tech giants warn Digital Markets Act is failing

Apple and Google have urged the European Union to revisit its Digital Markets Act, arguing the law is damaging users and businesses.

Apple said the rules have forced delays to new features for European customers, including live translation on AirPods and improvements to Apple Maps. It warned that competition requirements could weaken security and slow innovation without boosting the EU economy.

Google raised concerns that its search results must now prioritise intermediary travel sites, leading to higher costs for consumers and fewer direct sales for airlines and hotels. It added that AI services may arrive in Europe up to a year later than elsewhere.

Both firms stressed that enforcement should be more consistent and user-focused. The European Commission is reviewing the Act, with formal submissions under consideration.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot! 

Google and Flo Health settle health data privacy suit for $56 million

Google has agreed to pay $48 million, and Flo Health, a menstrual tracking app, has agreed to pay $8 million to resolve claims that the app shared users’ health data without their consent.

The lawsuit alleged that Flo used third-party tools to transmit personal information, including menstruation and pregnancy details, to companies like Google, Meta, and analytics firm Flurry.

The class-action case, filed in 2021 by plaintiff Erica Frasko and later consolidated with similar complaints, accused Flo of violating privacy laws by allowing user data to be intercepted via embedded software development kits (SDKs).

Google’s settlement, disclosed this week, covers users who inputted reproductive health data between November 2016 and February 2019.

While neither Flo nor Google admitted wrongdoing, the settlement avoids the uncertainty of a trial. A notice to claimants stated the resolution helps sidestep the costs and risks of prolonged litigation.

Meta, a co-defendant, opted to go to trial and was found liable in August for violating California’s Invasion of Privacy Act. A judge recently rejected Meta’s attempt to overturn that verdict.

According to The Record, the case has drawn significant attention from privacy advocates and the tech industry, highlighting the potential legal risks of data-sharing practices tied to ad-tracking technology.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!

Gemini’s image model powers Google’s new Mixboard platform

Google has launched Mixboard, an experimental AI tool designed to help users visually explore, refine, and expand ideas both textually and visually. The Gemini 2.5 Flash model powers the platform and is now available for free in beta for users in the United States.

Mixboard provides an open canvas where users can begin with pre-built templates or custom prompts to create project boards. It can be used for tasks such as planning events, home decoration, or organising inspirational images, presenting an overall mood for a project.

Users can upload their own images or generate new ones by describing what they want to see. The tool supports iterative editing, allowing minor tweaks or combining visuals into new compositions through Google’s Nano Banana image model.

Quick actions like regenerating and others like this enable users to explore variations with a single click. The tool also allows text generation based on context from images placed on the board, helping tie visuals to written ideas.

Google says Mixboard is part of its push to make Gemini more useful for creative work. Since the launch of Nano Banana in August, the Gemini app has overtaken ChatGPT to rank first in the US App Store.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!

Google expands Search Live to US users

Google has expanded its Search Live feature to all app users in the US after several months of testing.

The tool allows people to hold voice conversations with AI Mode inside Google Search and even share a live camera feed. With this, the system can interpret surroundings, respond in real time, and suggest web links for deeper exploration.

The feature, powered by a customised version of Google’s Gemini chatbot, can run in the background while other apps are open. Google highlighted uses ranging from travel help to troubleshooting tasks.

Search Live is currently available only in English in the US. It can be enabled in the Google app by tapping the new Live icon or through Google Lens by selecting the Live button at the bottom of the screen.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!