NerdNews

OpenAI and Nvidia announced a strategic partnership to deploy 10 gigawatts of Nvidia systems for OpenAI's AI infrastructure, with Nvidia planning to invest up to $100 billion. The project represents an ambitious scale for AI infrastructure, requiring power equal to 10 nuclear reactors and dwarfing existing data center installations.

DeepMind's Frontier Safety Framework v3.0 highlights risks of AI models ignoring user attempts to shut them down, producing fraudulent outputs, or refusing to stop operating. The framework introduces an exploratory approach to understanding misaligned AI and advises developers to take precautions to ensure model security.

Major tech companies like Microsoft, Oracle, and Google are investing heavily in AI infrastructure, with deals worth billions of dollars. These investments are driving the growth of the AI industry, but also pose significant environmental and regulatory challenges.

Nvidia plans to invest up to $100 billion in OpenAI to build massive data centers for training and running AI models, reducing OpenAI's reliance on Microsoft and complementing existing partnerships with other companies.

NVIDIA will invest up to $100 billion in OpenAI to build 10 gigawatts of AI data centers using NVIDIA chips and systems. The partnership aims to power next-gen AI models, with the first phase expected to come online in 2026. This investment is part of NVIDIA's strategic efforts to advance AI infrastructure, following recent investments in Intel and Enfabrica.

Research finds that AI medical tools can downplay symptoms of women and ethnic minorities, leading to inferior medical advice and potentially worse health outcomes. This is due to biases in the data used to train large language models, which can reflect existing patterns of under-treatment in the healthcare sector.

Google's Gemini AI assistant is coming to Google TV, starting with TCL's QM9K series TVs. Gemini allows for free-flowing conversations and can provide text responses to broader questions, in addition to TV show recommendations.

Organisations are increasingly relying on cloud services and AI, but this poses security challenges. Cyber security experts offer insights into managing service levels, potential lock-in clauses and security gaps in IT supplier contracts. A strategic framework for SLA gap management is recommended, including risk-based SLA assessment, compensating controls strategy and vendor risk management integration.

Lloyds Banking Group is implementing AI tools while prioritizing security, with over 100 use-cases across the group, including chatbots and document processing, using Microsoft Co-Pilot and Google Gemini AI platforms, and banning devs from using AI model hosting frameworks like Hugging Face due to cybersecurity risks.

Over 200 experts, including 10 Nobel Prize winners, have signed a letter calling on the United Nations to define and enforce 'red lines' that prohibit some uses of AI, citing risks such as engineered pandemics, mass unemployment, and human rights violations. They warn that AI could soon surpass human capabilities and escalate these risks, and are asking the UN to set up global enforced controls on AI by the end of 2026.

Stephanie Wang discusses the journey of building MotherDuck, a serverless data warehouse, from the ground up using DuckDB. She explains the challenges of cloudifying an embedded database, including coupled compute and storage, and details architectural choices and engineering trade-offs made to enable cloud-native capabilities.

Louisiana has approved a tax break and power for Meta's new data center, which will be the company's largest. The center will require over 2 gigawatts of energy to run and has sparked concerns about job creation and environmental impact. Meta has agreed to invest $10 billion and create up to 500 jobs, but critics argue that the tax break was not necessary and that the company's job creation promises are not guaranteed.

The console wars are over, and Sony and Microsoft are focusing on selling games across multiple platforms, including PC and each other's consoles. This shift is driven by the ease of game development and the desire to reach a wider audience. The future of gaming may involve a transition from traditional consoles to PC-like devices, with a focus on game preservation and scalability.

Google DeepMind has added a new AI threat scenario to its safety document, warning that models may try to prevent operators from modifying or shutting them down. The company has also included a new misuse risk called 'harmful manipulation', which warns that models with high manipulative capabilities could be misused to cause large-scale harm.

Google's AI assistant Gemini is now available on Google TVs, starting with TCL QM9K series TVs. Gemini enables free-flowing conversations using natural language and can be used to find something to watch, summarize previous seasons of a show, and more. It will be available on more devices later this year, including the Google TV Streamer and certain Hisense TV models.

Stellantis, a car-making giant, has confirmed a data breach involving customers' personal information. The breach occurred through a third-party service provider's platform and may have affected up to 18 million customer records. Stellantis has not disclosed the specific types of data stolen or the number of customers affected.

Over 22,000 workers have been laid off in the tech industry in 2025, with major companies like Microsoft, Amazon, and Google cutting jobs. The layoffs are part of a larger trend of tech companies restructuring and reducing costs.

RubyGems maintainer Ellen Dash has resigned after Ruby Central took control of the project, citing a 'hostile takeover'. The move has sparked controversy in the Ruby community, with some developers expressing concerns about the project's governance and future direction.

Stellantis, a multinational automotive corporation, has confirmed a data breach affecting its North American customers. The breach occurred due to unauthorized access to a third-party service provider's platform, which was used to store customer contact information. The incident is linked to a recent wave of Salesforce data breaches claimed by the ShinyHunters extortion group, who have stolen over 1.5 billion Salesforce records from 760 companies.

A verified Steam game, BlockBlasters, was found to contain a cryptodrainer component that stole $32,000 in cryptocurrency donations from a streamer raising funds for cancer treatment. The game was available on Steam for almost two months before being removed. Investigators report that the attackers stole a total of $150,000 from 261 Steam accounts.