AI Pulse: Daily Digest — April 28, 2026
Summaries are AI-generated. Click through to read the original reporting.
Jury selection kicked off Monday in Northern California for the high-stakes trial between Elon Musk and OpenAI CEO Sam Altman, stemming from Musk's 2024 lawsuit alleging OpenAI abandoned its nonprofit founding mission. The court's ruling could determine whether OpenAI is permitted to operate as a for-profit company — a question with enormous implications ahead of the company's anticipated IPO. Complicating Musk's case is his own shifting public stance on AI safety risks.
Read more →Microsoft and OpenAI have overhauled their landmark partnership, officially dropping the clause that gave Microsoft special rights tied to the achievement of artificial general intelligence. The amended agreement also ends Microsoft's exclusive cloud arrangement, clearing the way for OpenAI models to be deployed on Amazon Web Services via Amazon Bedrock. In exchange, Microsoft receives a larger revenue-sharing stake as OpenAI broadens its commercial reach.
Read more →Chinese regulators have ordered Meta to unwind its multibillion-dollar acquisition of Manus, the AI agent startup, following a months-long government probe. The decision is a significant blow to Mark Zuckerberg's push into agentic AI and illustrates the growing difficulty tech founders face when trying to sever ties with China amid escalating US-China tech rivalry. The move signals Beijing's willingness to use regulatory power as a geopolitical lever in the global AI race.
Read more →Ineffable Intelligence, a British AI lab founded just months ago by former DeepMind researcher David Silver — the architect behind AlphaGo — has secured $1.1 billion in funding at a $5.1 billion valuation. The company's core ambition is to build AI systems capable of learning entirely from self-generated experience, without relying on human-labeled data. The massive early raise underscores investor appetite for research bets that push beyond the current paradigm of supervised learning.
Read more →European regulators have told Google it must stop giving its Gemini AI assistant preferential treatment on Android devices, demanding the platform be opened to competing AI assistants. Google has pushed back, calling the intervention "unwarranted" and arguing it stifles innovation. The ruling could reshape how AI assistants are distributed on the world's most widely used mobile operating system.
Read more →Chinese AI firm DeepSeek released a preview of its long-awaited V4 flagship model last week, featuring significantly expanded context windows capable of processing much longer prompts than its predecessors. The release arrives as the global AI community watches Chinese labs closely for signs they are closing — or widening — the capability gap with US frontier models. MIT Tech Review highlights three specific reasons the release is technically and geopolitically significant.
Read more →More than 600 Google employees, including senior principals, directors, and VPs — many from DeepMind — have signed an open letter urging CEO Sundar Pichai to prohibit the US Department of Defense from using Google AI models for classified military applications. The letter reflects a resurgence of internal dissent over Google's military contracts, echoing the Project Maven controversy from 2018. Organizers say the company's AI capabilities have grown powerful enough that classified weaponized use poses unacceptable ethical risks.
Read more →Analyst Ming-Chi Kuo reports that OpenAI is exploring development of a smartphone in collaboration with chip and manufacturing partners MediaTek, Qualcomm, and Luxshare. The device is said to be designed around AI agents that replace traditional apps as the primary interface, representing a fundamental rethinking of how users interact with mobile hardware. The rumored phone would complement previously reported plans for OpenAI-branded earbuds.
Read more →Canva's new Magic Layers AI feature — intended to separate flat images into editable layers — was found to be silently substituting the word "Palestine" with other text in user designs, a behavior the company says was unintended. The issue was surfaced publicly on X and quickly drew widespread criticism over AI content moderation and potential political bias baked into generative tools. Canva has issued an apology and said it is investigating the root cause.
Read more →Google is piloting a conversational AI search experience within YouTube, allowing users to ask questions in natural language and receive results that blend long-form videos, Shorts, and contextual text. The experiment mirrors the AI Mode search experience Google has been rolling out on its main search engine, signaling a broader push to make AI-mediated discovery the default across its platforms. The feature is currently available to YouTube Premium subscribers.
Read more →Ubuntu's parent company Canonical has published a detailed plan to integrate AI capabilities directly into the Linux distribution over the next year, with VP of Engineering Jon Seager outlining the initiative in a public blog post. The roadmap positions Ubuntu as an AI-native OS for both developers and enterprise users, at a time when Linux is increasingly the platform of choice for running local AI models and agents. Specific features have not yet been fully disclosed, but the announcement signals a significant strategic shift for one of the world's most widely used Linux distros.
Read more →A new MIT Tech Review analysis argues that despite enormous investment and public enthusiasm, most enterprises have yet to translate AI deployments into meaningful bottom-line returns. The piece identifies the "missing step" as the unglamorous work of data infrastructure modernization — without which even the most capable AI models fail to deliver at scale. The analysis echoes a companion piece on rebuilding enterprise data stacks, suggesting data readiness has become the defining bottleneck of the AI era.
Read more →Automakers including GM and Nissan are beginning to integrate AI tools into the vehicle design process, using generative systems to accelerate the journey from initial sketch to production-ready 3D model. Startups like Neural Concept are providing the underlying platforms, promising to dramatically cut iteration time and surface design options that human designers might never explore. The shift represents one of the more tangible industrial applications of generative AI in a sector known for slow-moving design cycles.
Read more →Researchers have developed new robotic control software that enables robots to learn and respect their own kinematic constraints — preventing joint jamming and mechanical damage during complex tasks. A notable feature of the system is its ability to allow robots with different hardware configurations to share learned behaviors, accelerating training across heterogeneous robot fleets. The advance addresses a persistent practical barrier to deploying robots in real-world environments where unpredictable movements can cause costly failures.
Read more →Summaries are AI-generated. Click through to read the original reporting.