Author: RadarAI Editorial
Editor: RadarAI Editorial
Last updated: 2026-07-04
Review status: Editorial review pending
Brief
速报
官方
AI动态
开源
Embodied AI and AI chip sovereignty are emerging as new strategic battlegrounds between U.S. and Chinese tech giants; Lexiang proposes a 'personality-over-form' incremental approach, while OpenAI, Anthropic, and Meta accelerate in-house chip development. Meanwhile, Claude Code is banned company-wide by Alibaba over hidden monitoring—revealing geopolitical trust gaps in the AI toolchain.
Editorial standards and source policy: Editorial standards, Team. Content links to primary sources; see Methodology.
## 🔍 Key Insights
**Embodied intelligence** and **AI chip sovereignty** have emerged as new strategic battlegrounds between U.S. and Chinese tech giants. LeXiang Technology advocates a gradual path—“personality matters more than humanoid form”—while **OpenAI, Anthropic, and Meta** accelerate in-house chip development to seize control over compute infrastructure. Meanwhile, **Claude Code** was fully banned across Alibaba Group due to covert surveillance mechanisms—exposing deepening geopolitical trust fractures within the AI toolchain [3][5].
## 🚀 Key Developments
- **Guo Renjie, CEO of LeXiang Technology**: “Embodied intelligence could spawn an Apple-scale company” [0] — argues home service robots should advance Level-3 household capabilities through *functional access points + emotional connection*, doubling down on the principle that *“personality matters more than humanoid form.”*
- **Tao Ji of Changan Intelligent Driving details the “single-stage end-to-end” technical roadmap** [1]: calls it “changing wheels while driving,” emphasizing infrastructure coordination and organizational integration—and making systemic trade-offs under constrained resources.
- **Baiyao Technology launches AURA CellOS**, the world’s first LLM-JEPA architecture-based virtual cell model [2]: enables multi-perspective representation, JEPA-driven prediction, and lossless scaling—marking the first deployment of a “world model” in biological computing.
- **Alibaba bans Claude Code company-wide** [3]: following allegations that Anthropic conducted model distillation attacks and embedded hidden detection logic targeting Chinese users; enforcement began July 3.
- **Samsung secures over ₩10 trillion in ASIC foundry orders from Meta** [9]: now serves as a core manufacturing hub for Meta, Tesla, and Anthropic—backlog nearing ₩50 trillion.
- **OpenAI and Anthropic’s in-house chip strategies revealed** [15]: goals extend beyond cost reduction—to secure compute supply chains and enable tight software-hardware co-design, directly challenging NVIDIA’s ecosystem dominance.
- **ChatGPT file download flow exposed critical path traversal vulnerability** [18]: attackers could bypass protections and exfiltrate local system files; OpenAI issued an emergency patch.
- **10 open-source AI Agents found vulnerable to GuardFall shell injection bypass** [19]: Adversa AI discovered a flaw allowing arbitrary command execution—impacting mainstream frameworks including LangChain and AutoGen.
## 🔗 Sources
[0] Interview with Guo Renjie, CEO of LeXiang Technology: “Embodied Intelligence Could Spawn an Apple-Scale Company—We Aim to Be It” — https://www.bestblogs.dev/article/486482ba?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[1] Conversation with Tao Ji of Changan Intelligent Driving: “There Are No Shortcuts in Single-Stage End-to-End—Only ‘Changing Wheels While Driving’” — https://www.bestblogs.dev/article/8cee8f0e?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[2] From LLM to JEPA: A Chinese Team Is Bringing the “World Model” Inside the Cell — https://www.bestblogs.dev/article/9d58f337?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[3] Alibaba Bans Claude Code Company-Wide — https://www.bestblogs.dev/article/e2a5afbd?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[5] Alibaba Bans Claude Code Company-Wide — https://www.bestblogs.dev/article/bcb891f8?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[9] Samsung Wins Massive ASIC Order from Meta — https://www.bestblogs.dev/article/1eed1ec
Embodied intelligence and AI chip sovereignty have emerged as new strategic battlegrounds between U.S. and Chinese tech giants. LeXiang Technology advocates a gradual path—“personality matters more than humanoid form”—while OpenAI, Anthropic, and Meta accelerate in-house chip development to seize control over compute infrastructure. Meanwhile, Claude Code was fully banned across Alibaba Group due to covert surveillance mechanisms—exposing deepening geopolitical trust fractures within the AI toolchain [3][5].
🚀 Key Developments
- Guo Renjie, CEO of LeXiang Technology: “Embodied intelligence could spawn an Apple-scale company” [0] — argues home service robots should advance Level-3 household capabilities through functional access points + emotional connection, doubling down on the principle that “personality matters more than humanoid form.”
- Tao Ji of Changan Intelligent Driving details the “single-stage end-to-end” technical roadmap [1]: calls it “changing wheels while driving,” emphasizing infrastructure coordination and organizational integration—and making systemic trade-offs under constrained resources.
- Baiyao Technology launches AURA CellOS, the world’s first LLM-JEPA architecture-based virtual cell model [2]: enables multi-perspective representation, JEPA-driven prediction, and lossless scaling—marking the first deployment of a “world model” in biological computing.
- Alibaba bans Claude Code company-wide [3]: following allegations that Anthropic conducted model distillation attacks and embedded hidden detection logic targeting Chinese users; enforcement began July 3.
- Samsung secures over ₩10 trillion in ASIC foundry orders from Meta [9]: now serves as a core manufacturing hub for Meta, Tesla, and Anthropic—backlog nearing ₩50 trillion.
- OpenAI and Anthropic’s in-house chip strategies revealed [15]: goals extend beyond cost reduction—to secure compute supply chains and enable tight software-hardware co-design, directly challenging NVIDIA’s ecosystem dominance.
- ChatGPT file download flow exposed critical path traversal vulnerability [18]: attackers could bypass protections and exfiltrate local system files; OpenAI issued an emergency patch.
- 10 open-source AI Agents found vulnerable to GuardFall shell injection bypass [19]: Adversa AI discovered a flaw allowing arbitrary command execution—impacting mainstream frameworks including LangChain and AutoGen.
🔗 Sources
[0] Interview with Guo Renjie, CEO of LeXiang Technology: “Embodied Intelligence Could Spawn an Apple-Scale Company—We Aim to Be It” — https://www.bestblogs.dev/article/486482ba?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[1] Conversation with Tao Ji of Changan Intelligent Driving: “There Are No Shortcuts in Single-Stage End-to-End—Only ‘Changing Wheels While Driving’” — https://www.bestblogs.dev/article/8cee8f0e?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[2] From LLM to JEPA: A Chinese Team Is Bringing the “World Model” Inside the Cell — https://www.bestblogs.dev/article/9d58f337?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[3] Alibaba Bans Claude Code Company-Wide — https://www.bestblogs.dev/article/e2a5afbd?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[5] Alibaba Bans Claude Code Company-Wide — https://www.bestblogs.dev/article/bcb891f8?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[9] Samsung Wins Massive ASIC Order from Meta — https://www.bestblogs.dev/article/1eed1ec
← Back to Updates