☕️ OpenAI's Proposed Ban on DeepSeek, Google's Gemma 3 Release & Gemini's Next Major Step Into Robotics

OpenAI and Google continue to dominate the headlines on the regular, be it with proposed bans on Chinese-backed AI models or the release of AI models so small they can run on a phone or single GPU with more details below:

• OpenAI sets its sights on having the US ban AI models backed by the Chinese government in a recent proposal

• Google launches a family of some of the most capable lightweight AI models to date with the release of Gemma 3

• Google's AI research arm DeepMind brings Gemini to the world of robotics

Join us at AI Tangle as we untangle this week's happenings in AI!

OpenAI lays out a sudden proposal to ban PRC-backed AI models like DeepSeek over security concerns

​OpenAI has recently submitted a bold policy proposal as part of the Trump Administration's "AI Action Plan" to urge the US government to either restrict or outright ban AI models developed by organizations backed by the People's Republic of China (PRC). Unsurprisingly, OpenAI outed DeepSeek specifically, which it describes as "state-subsidized" and "state-controlled," due to security concerns over data sharing under Chinese law. OpenAI claims such models pose risks like privacy breaches, intellectual property theft, etc., calling for bans in countries classified as "Tier 1" under the Biden Administration's export rules.

​

Google releases its next Gemma 3 family of lightweight AI models that run on a single GPU

​Not long after Alibaba showed off its miniature model, Google unveiled the next-in-line series of Gemma lightweight AI models called Gemma 3. Coming in 4 different sizes ranging from 1 billion to 27 billion parameters, the Gemma 3 family can run on hardware anywhere from a smartphone to just a single enterprise Nvidia H100 GPU. Gemma 3 offers advanced multimodal capabilities and, for its size, a fairly large 128,000-token context window, letting it handle lengthy documents, extended video segments, or high-quality images along with additional automation features and function-calling.

​

DeepMind leverages Google's Gemini suite to develop wholly models custom-built for the robotics industry

​In a blog post released on Wednesday, Google DeepMind announced that it brought its breath of Gemini 2.0 models into the world of robotics, showcasing two new Gemini Robotics AI models as part of a demo. The demos showed robots equipped with Gemini Robotics to do physical stunts such as slam-dunking a miniature basketball or general tasks like adjusting fruit placement in a moving bowl. To further advance its efforts, the tech bigwig has partnered with companies like Apptronik, Agile Robots, and Boston Dynamics, as Google aims to build more versatile and physically capable robots.

​

Authors and publishers in France sue Meta over the use of copyrighted material in AI training sets

Authors and publishers in France have recently taken legal action against Facebook's parent company, Meta, over accusations of the company using their copyrighted works in AI training sets without permission. The National Publishing Union, the National Union of Authors and Composers, and the Societe des Gens de Lettres are the main three groups taking Meta to court in Paris. They collectively demand the removal of data directories created from their members' works, arguing that this unauthorized use constitutes copyright infringement and parasitism, and could lead to AI-generated "fake books" that compete with real content.

​

Recently submitted court documents reveal Google's interest in Anthropic's technology and potential

According to legal filings submitted by Google as part of an antitrust case, court documents revealed that the company owns a major stake in AI unicorn Anthropic - 14%, to be exact. Additionally, the tech giant is looking to invest another $750 million in September through a type of loan known as convertible debt, which would put the company's investment into Anthropic over $3 billion. However, despite Google's share in the company, the documents also state that Google owns no voting rights, no board observer rights, and no board seats within Anthropic.

​

Salesforce announces a $1 billion investment plan in Singapore's AI sector over the next five years

Earlier this Wednesday, Salesforce announced plans to invest $1 billion in Singapore over the next five years to speed up the country's digital transformation and help along with the adoption of its AI platform, Agentforce. The investment looks to tackle Singapore's labor shortages during a time of an aging population by putting the country in a spot "to build a vibrant hub for AI innovation," said Jermaine Loy, managing director of the Singapore Economic Development Board.

​

Microsoft brings its in-house AI to Notepad to help it write summaries for you

Microsoft has recently started a trial run of AI-powered summaries in Notepad on the Canary and Dev channels as part of an update for Windows Insiders. In its current state, users can simply highlight text, right-click to select a new "Summarize" option, or use the keyboard shortcut Ctrl + M to generate a summary with customizable length. Alongside this new feature, Microsoft is also testing options to view recently closed files and a "draw & hold" function in the Snipping Tool, which automatically straightens drawn shapes for those of us with shaky hands.

​Hoop - Made for the busy professional, Hoop is an AI task management platform that captures and prioritizes your tasks with a global task list across your teams.

​Nuelink - Nuelink helps you organize, automate, analyze, and manage your social media from one unified place, saving you time and enabling you to focus on your business while your social media runs itself.

​Roundtable - Clean your survey responses with Roundtable, an easy-to-integrate API with behavioral tracking, dynamic clustering, and more to cut down on time spent analyzing survey data.

Is The Future of Code AI-Generated? (3-min read)

From Anthropic to IBM to Meta, tech leaders across the industry aren't shy about sounding the alarm that the role and position of a software developer may radically change within just months from now. Though the exact timeframes and impact differ from opinion to opinion, the vast majority remain bullish on the shift - but why?