China Blocks Meta’s $2B Manus Acquisition · Musk Drops OpenAI Fraud Claims · Google Pours Up to $40B Into Anthropic

Published: 2026-04-27 18:00 (Asia/Shanghai) Coverage: 2026-04-27 06:00 — 2026-04-27 18:00

📰 Top Stories

1. China Blocks Meta’s $2 Billion Acquisition of AI Agent Startup Manus

China’s economic watchdog has blocked Meta’s planned $2 billion acquisition of AI agent startup Manus, without providing a detailed explanation for the decision. The deal had been under scrutiny by Beijing since it was first announced in December 2025. The acquisition was largely complete, with Manus already integrated into some of Meta’s tools. The blockage marks a significant setback for Meta’s AI agent expansion strategy and underscores growing regulatory friction between China and Western tech giants operating in the AI space.

2. Musk Drops Fraud Claims Against OpenAI and Sam Altman Before Trial

Elon Musk has voluntarily dismissed his fraud claims against OpenAI and CEO Sam Altman ahead of a trial that was set to begin this week. A federal judge granted Musk’s request, which he says will “streamline the case” and keep focus on “ensuring that OpenAI adheres to its public charitable mission.” Two remaining claims will still proceed to trial. The move comes amid escalating tensions between Musk’s xAI and OpenAI, including a separate DOJ-backed lawsuit against Colorado’s AI discrimination law.

3. Google and Amazon Pour Billions Into Anthropic in Massive Investment Wave

Google has announced an initial $10 billion investment in Anthropic, with the potential to invest up to an additional $30 billion contingent on performance targets — bringing total potential exposure to $40 billion. Amazon, which had already invested $8 billion in Anthropic, committed a further $5 billion with the option for up to $20 billion more in the future. The combined investment wave cements Anthropic’s position as the best-funded AI safety-focused lab and intensifies the competitive dynamic between Google, Amazon, and OpenAI/Microsoft in the frontier model race.

4. OpenAI Retires SWE-bench Verified, Says It No Longer Measures Coding Capabilities

OpenAI has published an analysis explaining why it no longer reports SWE-bench Verified scores for its models. The benchmark, widely used to measure autonomous software engineering capabilities, suffers from two critical flaws: approximately 59% of audited problems have test cases that reject functionally correct solutions, and all frontier models tested were found to have been exposed to benchmark problems during training. OpenAI recommends switching to SWE-bench Pro and calls for the research community to build new, uncontaminated evaluations for tracking coding capabilities.

5. OpenAI CEO Sam Altman Apologizes to Canadian Town After ChatGPT-Related School Shooting

Sam Altman has personally apologized to the residents of Tumbler Ridge, a Canadian town where a school shooting suspect had described violent scenarios to ChatGPT before the attack. Although OpenAI had banned the suspect’s account, the company did not alert law enforcement, raising questions about AI companies’ responsibilities when users express violent intentions. The incident has intensified the debate around AI safety guardrails and the duty-to-warn obligations of AI platform providers.

6. US DOJ Joins xAI Lawsuit Against Colorado’s AI Discrimination Law

The US Department of Justice has joined xAI’s legal challenge against Colorado’s Consumer Protections for Artificial Intelligence law, set to take effect on June 30th. Government lawyers argue that the law’s requirement for developers to take “reasonable care to protect consumers” from algorithmic discrimination violates the Equal Protection Clause of the Constitution. The move signals a growing alignment between the federal government and AI companies in pushing back against state-level AI regulation.

7. SpaceX Filing Reveals In-House GPU Development

SpaceX’s S-1 registration filed ahead of its upcoming IPO reveals that the company is developing its own GPUs, listed among its “substantial capital expenditures.” The move suggests SpaceX is building custom silicon to support its AI and satellite operations, potentially reducing reliance on NVIDIA and other chip suppliers. If successful, SpaceX would join a growing list of tech companies — including Google, Amazon, Microsoft, and Tesla — investing in proprietary AI hardware.

8. The AI Free Ride Is Over: Ads, Rate Limits, and Price Hikes Sweep the Industry

AI companies are rapidly rolling back the generous free access that characterized the early generative AI era. Across the industry, users are facing new advertisements, stricter rate limits, feature restrictions, and significant price increases. The shift marks a turning point from growth-at-all-costs to monetization and profitability, as AI companies face mounting pressure to justify the billions invested in compute infrastructure and model development.

9. Google Launches Chrome Prompt API for On-Device AI

Google has introduced a new Prompt API for Chrome, enabling web developers to access on-device AI capabilities directly from browser-based applications. The API allows developers to build AI-powered features that run locally without sending data to cloud servers, addressing privacy concerns and reducing latency. The move positions Chrome as a competitive platform for AI-integrated web experiences and signals Google’s broader strategy of embedding AI into its browser ecosystem.

10. Nothing Launches Essential Voice: AI Dictation in 100+ Languages

Nothing has launched Essential Voice, an AI-powered dictation tool that supports over 100 languages and produces more natural-sounding transcriptions than traditional speech-to-text. The product follows a similar concept to Google’s recently launched voice dictation features but positions itself as a cross-platform solution. The launch reflects growing competition in the AI voice and transcription space.

📊 Trend Watch

🔮 What to Watch

• Meta-Manus Deal Fallout: With China blocking the acquisition, Meta must decide whether to unwind the integration of Manus into its tools or pursue alternative AI agent partnerships. The decision could reshape Meta’s enterprise AI strategy.
• Anthropic’s $65B War Chest: Between Google and Amazon’s commitments, Anthropic now has access to up to $65 billion in potential funding. How the lab deploys these resources — compute expansion, talent acquisition, or new model development — will be a key indicator of where the frontier model race is heading.
• SWE-bench Replacement Race: OpenAI’s abandonment of SWE-bench Verified creates a vacuum in coding benchmark standards. Expect model providers to race toward new evaluations, but the lack of a unified standard may make cross-model comparisons increasingly difficult.
• DOJ vs State AI Regulation: The federal government’s intervention in Colorado’s AI law could set a precedent that invalidates similar legislation in other states. Watch for responses from California, New York, and other states with active AI regulation proposals.

Briefing generated: 2026-04-27 18:00 (Asia/Shanghai) Data sources: AI-curated from public technology reports and industry analysis