Edition: April 28, 2026

NVIDIA introduces Nemotron 3 Nano Omni, a long-context multimodal intelligence model for documents, audio, and video agents, offering best-in-class accuracy and efficiency in various benchmarks. The model extends the Nemotron multimodal line to a broader text, image, video, and audio model. It achieves top accuracy on several leaderboards, including MMlongbench-Doc, OCRBenchV2, and VoiceBench.

Over 3,000 software developers gathered at AI Dev 26 x SF to discuss the future of software development in the AI era, with industry leaders emphasizing the need for imagination, speed, and higher standards to drive innovation. The conference highlighted the transformative impact of AI on the tech industry, with experts sharing insights on the current state of play and the potential for AI to revolutionize software engineering. As the industry continues to evolve, developers are focusing on reducing defect rates and improving code correctness to unlock the full potential of AI.

Elon Musk testified in a federal courtroom that he co-founded OpenAI to prevent a catastrophic 'Terminator outcome' from artificial general intelligence, and that the company's shift towards for-profit goals has strayed from its original mission. Musk's lawsuit against OpenAI's CEO Sam Altman could result in significant governance changes and financial damages. The trial may impact OpenAI's plans for an initial public offering this year.

Amazon has begun offering new OpenAI products on its AWS platform, including the latest models and a code-writing service, following the end of Microsoft's exclusive rights to OpenAI products. This move marks the beginning of a deeper collaboration between AWS and OpenAI. The development comes after OpenAI's revised agreement with Microsoft, which allowed AWS to offer OpenAI's products.

OpenAI has partnered with Amazon to make its top models available on Amazon Web Services' Bedrock managed inference and agent platform, providing an alternative to accessing the models through OpenAI's APIs. This collaboration allows enterprises to build AI-augmented tools using OpenAI's models while addressing security and data privacy concerns. The move is part of OpenAI's promise to make its models available on AWS in exchange for up to $35 billion in new financing.

Elon Musk has taken the stand in a high-profile trial against OpenAI, a company he co-founded and later left due to disagreements over its direction. Musk is suing OpenAI CEO Sam Altman and president Greg Brockman, with the trial stemming from a soured relationship and competing business interests. The case marks the latest development in a long-standing feud between Musk and his former colleagues.

Amazon has introduced an AI-powered Q&A feature on product pages, allowing users to ask questions and receive conversational audio responses in real-time. The 'Join the chat' feature aims to provide key product details without requiring users to scroll through lengthy descriptions or reviews. This feature is part of Amazon's broader 'Hear the highlights' experience, offering short audio summaries on millions of product pages.

Hundreds of workers in Ireland tasked with training Meta's AI models are at risk of being laid off, with over 700 employees potentially losing their jobs due to reduced demand and operational requirements. The workers, employed by Dublin-based firm Covalen, were informed of the layoffs via a brief video meeting and were not allowed to ask questions. The layoffs are part of Meta's efforts to increase efficiency and reduce reliance on third-party vendors.

Google has granted the US Department of Defense access to its AI for classified networks, allowing for various uses, following Anthropic's refusal to do so without guardrails to prevent misuse. This deal marks Google's entry into a space where OpenAI and xAI have already signed agreements with the DoD. The move comes despite opposition from nearly 1,000 Google employees who signed an open letter urging caution.

IBM's AI coding assistant, Bob, has reached general availability, promising a 45% productivity boost for developers, with a premium package available for mainframe applications. The platform uses a mix of AI models to automate and augment the software development lifecycle. IBM claims Bob has delivered measurable gains, including 10x project-based ROI, in its own internal testing.

Amazon has unveiled a new AI-powered service called Copilot, aimed at simplifying work across various software applications, as well as an overhaul of its Amazon Connect service, which now includes four agentic AI components for different business needs. The Copilot service uses proactive alerts and always-on context to automate tasks, and Amazon Connect's new components focus on areas such as supply chains, hiring, healthcare, and customer experience. Amazon is entering a crowded market with these services, competing with established vendors like Microsoft, Salesforce, and Google.

Anthropic's AI chatbot Claude can now integrate with popular creative software such as Photoshop, Blender, and Ableton through new connectors, enabling users to access and manipulate data directly within these apps. This move marks the company's latest push into the creative industry. Claude's connectors can be used for specific functions in each app, streamlining creative workflows.

Google Translate is celebrating 20 years of breaking down language barriers with over 1 billion monthly users, now offering a new 'pronunciation practice' tool on Android for instant AI feedback. The tool supports languages like English, Spanish, and Hindi, helping users refine their speaking skills. Translate has come a long way since its inception, supporting about 250 languages today.

GitHub is shifting to a usage-based billing model for its Copilot AI service, aiming to align pricing with actual usage and ensure financial sustainability. The change, effective June 1, will charge users based on the specific AI tasks they perform, rather than a flat rate for broad categories of requests. This move reflects the varying backend computing costs associated with different AI tasks.

Elon Musk and OpenAI are heading to court in a trial that will delve into the early days of AI and potentially reveal secrets of key figures. The case will determine who deserves credit and compensation for AI developments. Musk's intentions and potential motivations for the trial are also being scrutinized.

YouTube is introducing an AI-powered search feature that provides guided answers with step-by-step results, including text and video, to help users find more relevant content. The 'Ask YouTube' feature is currently available to Premium subscribers in the US, with plans to expand to non-Premium users. This move is part of Google's broader push to integrate AI-powered search across its platforms.

Neurable, a BCI startup, is looking to license its 'mind-reading' technology to consumer wearables, allowing companies to integrate brain-sensing tech into products like headphones and headbands. The company's non-invasive technology uses EEG sensors and AI to analyze brain activity. Neurable aims to make its tech as ubiquitous as heart rate sensors on smartwatches.

Tenstorrent has announced the general availability of its Galaxy Blackhole AI compute platform, featuring 32 Blackhole accelerators in a 6U, $110,000 chassis, offering 23 petaFLOPS of dense FP8 performance. The system can be extended to support larger models and higher throughput by adding more systems. Tenstorrent claims its four-node Blackhole Galaxy Superclusters can process a 100,000 token prompt in less than four seconds.

The FIDO Alliance, with contributions from Google and Mastercard, is launching working groups to develop industry standards for validating and protecting transactions carried out by AI agents, aiming to prevent agent hijacking and ensure transparency. This effort seeks to establish a protective baseline for industries to adopt, promoting trust in agentic AI. The goal is to provide cryptographic proof of user authorization for agent-initiated transactions while maintaining user privacy.

Red Hat principal software engineer Sally O'Malley has released a new open source tool called Tank OS to make deploying and managing OpenClaw agents safer and easier. Tank OS loads OpenClaw onto a Linux OS in a secure container, allowing for safer and more controlled use of the AI agent. This development is significant as O'Malley is an OpenClaw maintainer, working to improve the project's enterprise use cases.

The European Commission has ordered Google to open up its Android platform to competing AI services, allowing rival assistants to access key device capabilities and interact with applications. The proposed measures aim to promote innovation and give users more choice in AI services. Google has disagreed, arguing that the AI market is already competitive and that the intervention would undermine privacy and security protections.

Elon Musk and OpenAI CEO Sam Altman are heading to court in a case that could determine the future of OpenAI and its potential to operate as a for-profit enterprise. The trial could have significant implications for the global AI race. Meanwhile, the AI industry is struggling to find a clear path to profitability, with many companies still figuring out how to turn their technology into tangible profits.

Google has reportedly signed a classified deal with the US Department of Defense, allowing the Pentagon to use its AI models for any lawful government purpose, despite opposition from Google employees. This agreement would place Google alongside other companies like OpenAI and xAI, which have also made similar deals with the US government. The deal has sparked concerns over the potential misuse of AI in harmful ways.

Top cybersecurity teams demonstrated their AI bug-finding systems at DARPA's Artificial Intelligence Cyber Challenge, identifying artificial and previously unknown bugs in software code. The event highlighted the capabilities of AI in cybersecurity, with some teams finding more than a dozen bugs that weren't intentionally inserted. This showcases the potential of AI in strengthening cybersecurity.

Vendor lock-in is becoming a significant issue for companies using AI models, with nearly 60% of organizations experiencing difficulties when trying to switch between AI platforms. This is due to technical dependencies and proprietary systems that make migration challenging. As a result, AI vendors are increasing their prices, with some models seeing price hikes of over 400%.

Bloomberg is revamping its Terminal with a chatbot-style interface called ASKB, which utilizes language models to help finance professionals condense labor-intensive tasks and test investment theses. The goal is to surface untapped insights in the vast dataset and make it easier for users to find key information. ASKB is currently in beta and available to about a third of the Terminal's 375,000 users.

China has blocked Meta's acquisition of AI startup Manus, citing laws and regulations, in a move that signals Beijing's desire to maintain control over domestic AI technology. The decision is a setback for Meta's AI ambitions, which have yet to deliver a consumer-facing AI product. The blockage may impact Meta's ability to expand its AI capabilities.

A jury has been selected for the trial of Musk v. Altman, with some jurors expressing concerns about Elon Musk and AI technology, but assuring the court they can remain impartial. The trial will determine whether OpenAI's nonprofit venture was improperly steered away from its original mission. The jury's verdict will be advisory, with the final decision resting with Judge Yvonne Gonzalez Rogers.

Microsoft's GitHub is shifting to a metered AI billing system for its Copilot customers, citing unsustainable costs under the current request-based model. The new system, which takes effect on June 1, will charge users based on their usage of AI services, with costs calculated in GitHub AI Credits. This change reflects a broader industry trend of companies reevaluating their AI pricing strategies in response to rising demand and costs.

NVIDIA and Siemens Healthineers have developed an AI-powered ultrasound imaging system called NV-Raw2Insights-US, which learns directly from raw sensor data to create personalized maps of sound speed for each patient, enabling real-time adaptive image focusing. This technology has the potential to revolutionize medical imaging by providing more accurate and detailed images. The system is deployed using NVIDIA Holoscan, an edge AI sensor processing platform.

Google is testing an AI-powered chatbot search feature for YouTube, allowing users to search in a conversational manner. The feature, called 'Ask YouTube', is currently available to YouTube Premium subscribers in the US. It provides results in the form of longform videos, YouTube Shorts, and text.

Claude, an AI tool, is being integrated into various creative software to aid professionals in their work, allowing for faster and more ambitious ideation and handling repetitive tasks. This integration is made possible through connectors with popular tools like Blender, Adobe, and Autodesk. The goal is to expand the creative possibilities for professionals, enabling them to take on larger-scale projects.

India's top tech services companies, including Infosys, HCL, TCS, and Wipro, are experiencing 'AI deflation', which is putting downward pressure on revenue despite overall growth. The companies are adapting to this change by leveraging AI to improve productivity and win new deals. Headcounts are mostly holding up, with minimal job cuts.

OpenAI has explicitly instructed its Codex model to avoid mentioning mythical creatures like goblins unless relevant to the user's query, sparking a meme after users reported the model's tendency to fixate on such topics. The phenomenon has been observed in OpenClaw, a tool that lets AI control computer apps. OpenAI's move highlights the unpredictable nature of AI models.

Elon Musk testified in a high-profile trial, positioning himself as a savior who wants to save humanity. He shared his background, from his childhood in South Africa to his current companies. Musk's testimony aims to establish his motivations and character in the lawsuit against OpenAI co-founder Sam Altman.

Taylor Swift is taking legal action to protect herself from AI copycats by filing trademark applications for phrases spoken by her, including 'Hey, it's Taylor Swift' and 'Hey, it's Taylor'. The move aims to prevent AI-generated content from misusing her voice and likeness. Swift's efforts may face complications due to the intersection of law and technology.

A vintage chatbot named Talkie has been created with training data limited to 1930, aiming to help understand AI thought processes. This chatbot's limited knowledge may provide a unique perspective on AI development. Its creators hope it will offer insights into AI thinking without the influence of modern biases.

Lovable has launched its vibe-coding app on iOS and Android, allowing users to build apps on the go via voice or text AI prompts. The app enables seamless switching between devices and receives notifications when a build is ready for review. The launch comes after Apple's recent crackdown on vibe-coding apps, with Lovable complying with the new rules.

Otter, an AI meeting notetaker app, has introduced a new feature that enables users to search across their enterprise tools, connecting to apps like Gmail, Google Drive, and Salesforce. This move aims to transform Otter into a full workspace where users can bring in data from different sources and make decisions. The company has also redesigned its AI assistant to be more consistently present and contextually aware.

Jury selection has begun in the lawsuit between Elon Musk and Sam Altman, with many prospective jurors expressing strong negative opinions about Musk. Some have called him 'greedy' and a 'world-class jerk' in their questionnaires. The case centers around alleged broken promises at OpenAI.

Elon Musk appeared unprepared and petty during his testimony in the Musk v. Altman trial, often diverting attention to his own accomplishments. His demeanor was notably different from his previous court appearance, where he was more charming and persuasive. Musk's testimony focused more on himself than the case against Sam Altman.