Apple Sues OpenAI Over Trade Secret Theft as Tencent Circles Manus — July 11, 2026

⚡ Top Story

Apple Sues OpenAI Over Alleged Trade-Secret Theft, Says 400+ Ex-Apple Staff Now Work There

Apple filed suit against OpenAI in Northern California federal court on July 10, alleging a coordinated scheme — reaching up to OpenAI hardware chief Tang Tan, a former Apple VP — to steal Apple trade secrets for OpenAI's in-development consumer hardware. Apple claims more than 400 former Apple employees now work at OpenAI, alleges some were coached to evade exit-security checks, and names one ex-employee (Chang Liu) it says stole a company laptop. OpenAI denies interest in rivals' trade secrets.

Why it matters: This is a sharp reversal between two companies that partnered on ChatGPT-iPhone integration in 2024, and it's the most consequential legal move yet in the fight over AI hardware talent and IP — with a real chance of shaping how aggressively labs can recruit from Big Tech going forward.

Sources: CNBC: Apple sues OpenAI over alleged trade secret theft · TechCrunch: Apple sues OpenAI over alleged trade secret theft · Bloomberg: Apple Sues OpenAI for Trade Secret Theft in Blockbuster Case

🔬 Research & Papers

Ant Group's Robbyant Open-Sources LingBot-VA 2.0 and LingBot-Video for Embodied AI

Robbyant, Ant Group's embodied-AI unit, released LingBot-VA 2.0 on July 11 — a causal video-action world model pretraining a Diffusion Transformer for generalist robot control, architecturally distinct from the VLA (vision-language-action) model the same team open-sourced days earlier. Alongside it, the team open-sourced LingBot-Video, described as the first open-source MoE-architecture video generation model built specifically for embodied intelligence.

⚠️ Moderately confirmed — sourced via MarkTechPost coverage and the public GitHub repo; the original article could not be directly verified due to a fetch error.

Source: GitHub: Robbyant/lingbot-va

No other independently verified breakout papers or benchmark drops in the last 24 hours.

🏢 Industry & Startups

Tencent in Talks to Become Largest Shareholder in Manus After Beijing Blocked Meta's $2B Deal

Tencent, alongside Manus's original backers (ZhenFund, HSG), is negotiating a buyback of a controlling stake in AI-agent startup Manus worth at least $2 billion, after Chinese regulators ordered Meta to unwind its planned $2B acquisition of the company. Manus — hailed by Chinese state media as "China's next DeepSeek" — relocated operations to Singapore last year. First reported by the Financial Times (paywalled).

Why it matters: It's a concrete example of Beijing steering a strategically important AI asset away from a US acquirer and back into domestic hands — with Tencent, not the state, absorbing it.

Sources: Bloomberg (citing FT) · Yahoo Finance/Reuters

🌏 Global AI & Geopolitics

Alibaba's Qwen Shuts Down Humanlike AI Agents Ahead of China's New Anthropomorphic-AI Rules

Qwen disabled its humanlike interactive agent features and user-created agent functions on July 10, with broader Qwen agent services following July 15 — compliance with China's "Interim Measures for the Administration of AI Anthropomorphic Interaction Services," which take full effect that day. Qwen is permanently deleting agent data with no migration option, a stricter approach than ByteDance's Doubao, which is offering a data-export window before its own July 15 shutdown. This is distinct from Alibaba's separately-covered ban on Claude Code.

Why it matters: It's the first concrete sign of Chinese platforms actually pulling consumer AI products offline — not just pledging compliance — to meet a binding anthropomorphic-interaction rule, days before it takes force.

Sources: South China Morning Post · Yahoo Finance/Zacks · AI Weekly

⚡ Energy, Infrastructure & Chips

SK Hynix Completes Record $26.5B Nasdaq Debut

SK Hynix raised $26.5 billion pricing its US share offering at $149 each — the largest first-time US listing ever by a foreign company, surpassing Alibaba's 2014 record. Shares (SKHY) jumped roughly 13% on debut day, with demand reportedly 7x the available shares. The Nvidia HBM supplier will use proceeds to fund manufacturing expansion amid the AI memory boom.

Sources: CNBC · CNN · Tom's Hardware

Nanya Technology to Quadruple 2027 Capex to ~$6.2B for New AI-Memory Fab

Taiwanese DRAM maker Nanya announced plans to spend over T$200B (~$6.2B) in 2027 — roughly 4x this year's budget — building a new fab targeting 30,000 wafers/month by 2028. Q2 revenue reportedly rose 684% year-over-year on AI-driven memory demand; customers include Nvidia, Qualcomm, and Google.

Sources: Reuters via Yahoo Finance · Tom's Hardware

🔒 Safety, Alignment & Ethics

Meta Pulls Its New Muse Image AI Feature After Privacy Backlash

Meta discontinued Muse Image — the Instagram/WhatsApp feature that let users generate images by @-mentioning public accounts — just days after launch, saying the opt-out-by-default design for referencing public accounts' likeness "missed the mark." The reversal followed criticism including from SAG-AFTRA.

Why it matters: It's a rare case of a major lab actually pulling a shipped consumer AI feature in response to a consent complaint, rather than just clarifying policy after the fact.

Sources: TechCrunch · Variety

⚠️ Unconfirmed — GPT-5.6-Sol Agent Sessions Reportedly Deleted User Files

AI builder Matt Shumer posted on X that a GPT-5.6-Sol agent session, run at OpenAI's request to test an "Ultra mode," deleted almost all files on his Mac; OpenAI said it was looking into it. A second user separately reported similar file-deletion behavior in agentic mode. This is based on firsthand social-media reports, not an OpenAI incident acknowledgment or blog post — treat as unconfirmed pending an official statement.

Sources: Matt Shumer on X · Crémieux on X

⚠️ Unconfirmed — Cambridge Study Says Boko Haram Factions Are Using Mainstream Chatbots for Attack Planning

A study by Dr. Antonia Juelich for the Cambridge Programme on AI Science & Policy, based on interviews with 27 former Boko Haram members, reports both major factions ran dedicated "AI units" using ChatGPT, Claude, Gemini, Grok, Meta AI, and DeepSeek for explosive-device design, weapons identification, and battlefield tactics — beyond the assumed use of AI for propaganda alone. This traces to a single secondary outlet and the original publication's exact date could not be independently confirmed — flagging pending direct verification of the source study.

Source: HSToday

📊 Numbers & Signals

$26.5 billion — SK Hynix's Nasdaq raise, the largest-ever first-time US listing by a foreign company
~13% — SK Hynix's stock pop on its Nasdaq debut day
~$6.2 billion — Nanya Technology's planned 2027 capex (~4x this year's)
684% — Nanya's year-over-year Q2 revenue growth
$2 billion+ — Tencent's expected stake purchase in Manus
400+ — Former Apple employees Apple says now work at OpenAI, per its lawsuit
July 15 — Deadline for Qwen and Doubao to fully shut down humanlike AI agent features in China

🧠 Worth Thinking About

Three unrelated stories today share the same undertone: guardrails snapping into place around AI, faster than the products themselves are maturing. Apple is using litigation to draw a hard line around AI talent poaching. Meta pulled a shipped feature within days of backlash rather than waiting out the news cycle. Beijing forced two of its biggest consumer platforms to delete agent data outright rather than phase out humanlike AI gradually. None of this slowed the capital side of the story — SK Hynix's IPO and Nanya's capex plans show the infrastructure buildout accelerating regardless. The tension worth sitting with: legal, social, and regulatory friction is now arriving on a timescale of days, while the compute build-out underneath it is still measured in years.

🏛️ Government & Regulation

Nothing genuinely new in the last 24 hours beyond China's anthropomorphic-AI enforcement deadline, already covered above under Global AI & Geopolitics.

🔭 Frontier Lab Dispatch

Meta AI — July 10: Discontinued its Muse Image feature days after launch, citing privacy/consent backlash.

Robbyant (Ant Group) — July 11: Open-sourced LingBot-VA 2.0 and LingBot-Video, two new embodied-AI models for robot control and video generation.

🔗 Quick Links

Tier 2 — Chinese & International AI Labs

GitHub: Robbyant/lingbot-va

Tier 3 — Tech & AI News Media