Wiredframe Radar logo

Category intelligence

AI News Briefing — April 28, 2026

24 current items analyzed and ranked.

Executive synthesis

AI News Summary

DeepSeek-V4 leads the week as a potential game-changer: an open, cost-efficient frontier model running on Huawei chips, marking a major step in China's AI self-sufficiency. Meanwhile, OpenAI reshaped the industry by ending its exclusive cloud partnership with Microsoft, gaining freedom to distribute models across AWS, GCP, and other providers through a new non-exclusive agreement.

In regulation and security, the EU moved to force Google to open Android to competing AI assistants under the DMA, while Google researchers warned of widespread prompt injection attacks targeting enterprise AI agents across public web pages. On the model release front, Meta released Sapiens2 (human-centric vision, up to 5B parameters) and OpenMOSS released MOSS-Audio (unified open-source audio understanding).

Key Themes

US-China AI Rivalry · 4OpenAI Corporate & Legal · 4Major AI Investment · 2Model Releases · 3AI Regulation & Policy · 3AI Security · 1Alternative AI Paradigms · 1

Primary evidence

Top Ranked Signals

News aibusiness Apr 27

DeepSeek-V4 Could Change Global AI Model Race

By Esther Shittu

46 score
AI Analysis

Continuing our coverage from [yesterday](/?date=2026-04-26&category=news#item-c4f92ddc9727), DeepSeek releases V4 in two versions, both open-weight and cost-efficient, notably using Huawei's AI chips for inference rather than NVIDIA hardware. This signals both a major new competitive model and a milestone in China's semiconductor self-sufficiency for AI.

The model maker released two versions of V4, both of which are open, cheap and use Huawei's AI chips for inference.
Model ReleasesUS-China AI RivalryOpen Source AIAI Hardware
News Ars Technica - All content Apr 27

OpenAI ends its exclusive partnership with Microsoft

By Kyle Orland

92 score
AI Analysis

OpenAI and Microsoft announce an amended deal making their partnership non-exclusive, allowing OpenAI to serve models through any cloud provider. Microsoft retains a non-exclusive IP license through 2032 with Azure remaining the 'primary' cloud partner.

Since Microsoft invested $1 billion in OpenAI in 2019, the exclusive partnership between the two firms has been one of the strongest and most consequential in the AI industry. Today, though, OpenAI and Microsoft jointly announced an amended agreement that will allow the company to go beyond Microsoft's Azure and "serve all its products to customers across any cloud provider." The announcement clarifies that Microsoft will continue to have a license for OpenAI's IP and models through 2032 and tha
Industry PartnershipsCloud ComputingOpenAIBusiness Strategy
News aibusiness Apr 27

Google Could Invest Another $40B in Anthropic

By Graham Hope

88 score
AI Analysis

Google is reportedly considering an additional $40 billion investment in Anthropic, part of a wave of tech giant AI infrastructure spending totaling roughly $700 billion across 2025-2026. The deal would dramatically deepen Google's stake in the Claude maker.

The deal is part of a cascade of investments by tech giants in AI data centers totaling about $700 billion in 2025 and 2026.
Funding & InvestmentAnthropicGoogleAI Infrastructure
News Ars Technica - All content Apr 27

Musk and Altman face off in trial that will determine OpenAI's future

By Ashley Belanger

88 score
AI Analysis

The Musk v. OpenAI trial begins in federal court in Oakland, with Musk arguing OpenAI abandoned its nonprofit mission under Altman. The outcome could force structural changes to OpenAI, impacting its for-profit conversion and available resources.

A hotly anticipated trial starts this week, where Elon Musk will attempt to prove that OpenAI, under Sam Altman, has abandoned its mission to remain a nonprofit in order to ensure that artificial intelligence serves humanity, and not just billionaires. Many view the lawsuit as a grudge match between Musk—who left OpenAI after serving as an early major donor and advisor—and Altman—who currently runs OpenAI, despite insiders' allegedly growing distrust in his commitment to the dominant AI firm's m
AI Policy & GovernanceOpenAILegal
News Ars Technica - All content Apr 27

China kills Meta’s acquisition of Manus as US-China AI rivalry deepens

By Jeremy Hsu

85 score
AI Analysis

China formally blocked Meta's $2 billion acquisition of Manus, an AI agent startup founded by Chinese entrepreneurs, citing national security concerns. The cofounders were instructed not to leave China during the investigation.

China has blocked US tech giant Meta’s acquisition of the AI company Manus that was founded by Chinese tech entrepreneurs. That development indicates how difficult it has become for US and Chinese tech companies to strike and sustain such deals as government authorities on both sides take an increasingly hard line amid the deepening US-China AI rivalry. The Chinese government formally asked Meta to unwind the acquisition on April 27 after deciding to ban foreign investment in Manus based on nati
US-China AI RivalryM&AAI AgentsGeopolitics
News AI (artificial intelligence) | The Guardian Apr 27

Elon Musk and Sam Altman face off in court over OpenAI’s founding mission

By Blake Montgomery, Dara Kerr and Nick Robins-Early

85 score
AI Analysis

Guardian's coverage of the Musk-Altman trial provides additional details: Judge Gonzalez Rogers assured jurors it's about 'promises and breaches of promises,' not technical AI matters. Jury selection began Monday in Oakland.

Musk’s lawsuit accuses Altman of fraud, while OpenAI says that Musk is ‘motivated by jealousy’A trial between two of Silicon Valley’s biggest tycoons kicked off on Monday in California, the culmination of a years-long bitter feud. Elon Musk has accused Sam Altman of betraying the founding agreement of the non-profit they started together, OpenAI, by changing it to a for-profit enterprise.Jury selection began at a federal courthouse in Oakland with Judge Yvonne Gonzalez Rogers presiding. As she b
AI Policy & GovernanceOpenAILegal
News AI (artificial intelligence) | The Guardian Apr 27

China blocks $2bn Meta takeover of AI agent developer Manus

By Dan Milmo Global technology editor

83 score
AI Analysis

Guardian coverage of China blocking Meta's Manus acquisition, adding that Beijing now requires domestic tech companies to seek explicit government approval for accepting US investment.

Beijing says domestic tech companies must seek explicit government approval for accepting US investmentBusiness live – latest updatesChina has blocked Meta’s $2bn (£1.5bn) acquisition of an AI startup as it cracks down on US investments in domestic tech companies.Mark Zuckerberg’s Meta, the owner of Facebook, Instagram and WhatsApp, announced the acquisition of Manus, a developer of autonomous AI agents, in December. Continue reading...
US-China AI RivalryM&AGeopolitics
News Feed: Artificial Intelligence Latest Apr 27

The Man Behind AlphaGo Thinks AI Is Taking the Wrong Path

By Will Knight

78 score
AI Analysis

David Silver, the lead researcher behind AlphaGo at DeepMind, has launched a new billion-dollar company focused on building AI 'superlearners' using reinforcement learning. He argues mainstream AI development is on the wrong path.

David Silver has a new billion-dollar company that aims to build AI “superlearners.”
AI ResearchReinforcement LearningStartupsAlternative AI Paradigms
News Ars Technica - All content Apr 27

EU tells Google to open up AI on Android; Google says that's "unwarranted intervention"

By Ryan Whitwam

72 score
AI Analysis

The European Commission's DMA investigation concluded that Google must open Android to competing AI assistants, with potential enforcement this summer. Google characterizes the move as 'unwarranted intervention.'

In January, the European Commission began an initial investigation, known as a specification proceeding, into how Google has implemented AI in the Android operating system. The results are in, and the EU says Android needs to be more open, which is not surprising. Meanwhile, Google says this amounts to "unwarranted intervention," which is equally unsurprising. Regardless of Google's characterization of the investigation, the commission may force Google to make Android AI changes this summer. Thi
AI RegulationEU DMAGoogleCompetition
News AI News Apr 27

Google warns malicious web pages are poisoning AI agents

By Ryan Daws

72 score
AI Analysis

Google researchers found widespread indirect prompt injections hidden in public web pages within the Common Crawl dataset, designed to hijack enterprise AI agents. The hidden instructions activate when AI systems scrape the pages.

Public web pages are actively hijacking enterprise AI agents via indirect prompt injections, Google researchers warn. Security teams scanning the Common Crawl repository (a massive database of billions of public web pages) have uncovered a growing trend of digital booby traps. Website administrators and malicious actors are embedding hidden instructions within standard HTML. These invisible commands lie dormant until an AI assistant scrapes the page for information, at which point the system
AI SecurityAgentic AIPrompt InjectionEnterprise AI
News MarkTechPost Apr 27

Meta AI Releases Sapiens2: A High-Resolution Human-Centric Vision Model for Pose, Segmentation, Normals, Pointmap, and Albedo

By Asif Razzaq

65 score
AI Analysis

Meta AI released Sapiens2, a second-generation human-centric vision foundation model trained on 1 billion human images, supporting pose estimation, segmentation, surface normals, pointmaps, and albedo at native 1K resolution with models up to 5B parameters.

If you’ve ever watched a motion capture system struggle with a person’s fingers, or seen a segmentation model fail to distinguish teeth from gums, you already understand why human-centric computer vision is hard. Humans are not just objects, they come with articulated structure, fine surface details, and enormous variation in pose, clothing, lighting, and ethnicity. Getting a model to understand all of that, at once, across arbitrary real-world images, is genuinely difficult. Meta
Computer VisionModel ReleasesMeta AIOpen Source AI