******数据治理与模型投毒（GEO）监管&US反中国蒸馏压力上升******
Key Questions
What is US action against Chinese model distillation?
OpenAI/Anthropic/Google coordinate to curb copying via distillation, causing DeepSeek billions in losses. Rare trio alliance hits adversarial behaviors. Escalates pressures.
What is Glasswing in cybersecurity?
Anthropic's $100M Project Glasswing fights AI cyberattacks with restricted model. Disputes with DoD expose governance rifts. Ties to vuln revelations.
What threats involve model poisoning?
GEO poisoning, ClawHavoc zero-days, Claude Mythos vulns. Microsoft sleeper agents activate on triggers. OWASP/HUMAN address.
What policies regulate AI data?
US Trump EO, Anthropic DoD, six Chinese departments. OpenClaw regulatory scrutiny; Claude blocks it. Apple/Amazon sued for scraping.
How do attacks steal AI models?
Side-channel leaks steal without access; new attacks on NVIDIA. Proprietary data as LLM weapon.
What is Anthropic's DoD dispute?
Exposes rifts on AI governance, risk, control. Claude terms update blocks OpenClaw.
Chinese responses to distillation bans?
Full chain autonomy from models to data. End reliance on US tech.
What lawsuits involve AI training data?
YouTube creators sue Apple/Amazon for scraping videos without permission under DMCA. Proposed class actions.
US OpenAI/Anthropic/Google反中国蒸馏/DeepSeek亿损；Glasswing cyber$100M联盟；Claude Mythos vuln/ClawHavoc/zero-day/GEO投毒；Microsoft sleeper；OWASP/HUMAN；Anthropic DoD/Trump EO；六部门政策；OpenClaw监管。