Ship Now, Patch Later: A Chinese Humanoid Crosses Into Europe With Its Backdoor Still Open

Capability & Integration

• Apple sued OpenAI on July 10, accusing hardware chief Tang Tan — a 24-year Apple veteran — and a former Apple engineer of a “coordinated campaign” to obtain confidential iPhone and Apple Watch designs ahead of OpenAI’s push into consumer devices; OpenAI pushed back on July 14, saying it is “not aware of any evidence that this complaint has merit,” per Bloomberg and TechCrunch.

• DeepSeek moved V4 out of preview into general availability this week, retiring the older deepseek-chat/deepseek-reasoner API aliases in favor of V4-Flash and introducing peak-hour pricing that doubles rates during Beijing business hours, per TechNode.

  Review

  “DeepSeek moved V4 out of preview into general availability this week… per TechNode” — the cited TechNode article is dated June 30 and only announces a planned mid-July launch; it can’t itself confirm GA actually shipped this week — verify against a source dated closer to the actual GA event.

  Unverified

  Claims that V4 approaches Opus-4.8-level performance at a fraction of the price originate with early testers and DeepSeek’s own marketing, not independent benchmarking.

Robotics

• Unitree’s H1 Pro began commercial sales in Europe on July 22 — the first Chinese-made humanoid to enter a Western commercial market — days after US lawmakers and the Pentagon flagged that every unit carries obligations under China’s National Intelligence Law to cooperate with state intelligence on request, per Tech Times. A Bluetooth exploit disclosed last September, using an AES key hardcoded identically across every G1, H1, R1, Go2, and B2 unit, still allows root-level takeover and remains unpatched, per IEEE Spectrum. This is a real commercial deployment, not a demo.

  Review

  “hardcoded identically across every G1, H1, R1, Go2, and B2 unit” — no source (IEEE Spectrum or independent write-ups on CVE-2025-60251) mentions an “R1” model in this exploit’s scope, only Go2/G1/H1/B2 — fix: verify against source text and remove R1 if unsupported.

  Review

  “still allows root-level takeover and remains unpatched” — stated as flat fact from a single source, but Unitree issued a Sept. 29, 2025 statement claiming most vulnerabilities were patched, disputed by independent researchers — add unverified flag noting the vendor/researcher dispute.

• 1X’s NEO factory in Hayward, California opened in April with capacity for 10,000 units this year, but 1X says NEO currently completes household tasks autonomously only 60–70% of the time, with the rest handled by remote human “Expert Mode” operators — a figure it projects to reach 95%+ by 2028, per The Robot Report.

  Unverified

  The autonomy percentage is 1X’s own figure; no independent measurement was found.

Hardware & Supply Chain

• TSMC pledged another $100 billion for four additional advanced fabs and packaging facilities in Arizona on July 16, bringing its total US commitment to $265 billion, per the Arizona Commerce Authority.
• Memory makers Micron, Samsung, and SK Hynix have all fallen more than 20% from recent highs into a bear market despite record results, the same week Intel reported a 59% year-over-year jump in Data Center and AI revenue — investors aren’t doubting that AI chip demand is real, they’re doubting it can keep growing fast enough to justify current valuations, per The Motley Fool and Yahoo Finance. It sits oddly next to the DRAM price-fixing allegations covered here yesterday: the firms accused of squeezing supply are the ones Wall Street doubts can keep the resulting margins.

Environmental & Cultural Impact

• Global data centers’ water footprint is projected to roughly double from 4.5 trillion liters in 2025 to 9.3 trillion liters by 2030, with electricity use nearly doubling to 945 TWh and AI accounting for about 40% of that draw, per a June 2026 report from the UN University Institute for Water, Environment and Health, via EurekAlert.
• 71% of Americans oppose a data center being built in their area — more than the 53% who oppose a local nuclear plant — with water use and energy use cited about equally among opponents, per Gallup.

AI in the Wild

A video purporting to show a Komodo dragon attack went viral on Instagram and TikTok this week; three independent AI-detection tools rated it 99–99.9% likely generative AI, per Lead Stories. Unlike the political fakes covered here recently, this one was built purely for engagement — most synthetic content circulating right now isn’t propaganda, it’s bait.

Takeaway