注册并分享邀请链接,可获得视频播放与邀请奖励。

Leo Underwood
@VesselGhost1984
National award-winning book publisher, editor & author. Digging into the money and influence behind the campaign against Jensen Huang.Go Broncos.AkA 梁一梦.
加入 June 2011
844 正在关注    7.7K 粉丝
Why does Anthropic hate open source so much it bankrolls D.C. to kill it? Huge thanks to former Supreme Court clerk @MarkZJia for helping me spot the coordinated witch hunt against Jensen Huang: nothing exposes a narrative faster than watching a top clerk flunk day-one 1L constitutional law—and then do it all over again to protect closed models. Watch his latest performance. Calling a government "naming and shaming" label a "balance" to "preserve open source competition" is pure D.C. doublespeak. Not a ban. A label. The target: US companies built on Chinese open models. Reuters' own reporter admitted the real effect: it "could actually have sort of a naming and shaming effect, without actually creating any new restrictions." So what does the label do? It doesn't inform. It scares your customers. No vote. No trial. Just a label and a smile. What does Georgetown law professor Mark Jia add? Read him slowly. He says a disclosure regime would "strike a balance between preserving open source competition and discouraging reliance on foreign models." Try that sentence on anything else: a balance between keeping your diner open and discouraging people from eating there. That's not a balance. That's a brake. Next he says neutrally worded disclosures are "generally fine from a legal perspective." Translation: green light. Then the only problem he can find: a label "worded in a manner that stigmatizes." Hold on. Hold on. That's the whole trick. Under the First Amendment, the government can compel only purely factual, uncontroversial disclosures—the kind that prevent consumer fraud. A politically loaded "naming and shaming" label is neither factual neutrality nor uncontroversial; it's compelled stigma. Reuters' reporter admits the effect is shame. The professor says: just don't let the words show it. He didn't remove the shame. He changed its name. He didn't fight the blacklist—he proofread it, coaching regulators on how to strip out the blatant buzzwords so their scarlet letter survives a court challenge and still does maximum commercial damage. The Constitution is a shield against compelled speech, not a style guide for government hit lists. The experiment. An honest rule gives the same answer when you swap the name. So swap it. Put ANTHROPIC under the label. Feb 27: The Pentagon branded Anthropic a "Supply-Chain Risk to National Security." Dario called it "very punitive and inappropriate." Next day, the professor went to bat: he tweeted the government was "treating the latter worse than Chinese AI firms, most courts would say: arbitrary and capricious," and even cited the Xiaomi and Luokung court victories proving the government can't slap arbitrary security labels on companies. Anthropic co-founder Chris Olah thanked MarkZJia by name for "independent legal commentary." For Anthropic: he cites Xiaomi's court win to strike down a government label. For US startups using open models: he proofreads the label so it survives court. Put JENSEN HUANG under the label. This month the professor amplified a post claiming Trump "KNEW Jensen Huang was sleeping with a Chinese spy." He floated blackmail. He listed the FBI, Commerce, the SEC with No EVIDENCE. Rule for a file: be neutral. Rule for a man: amplify a smear. That's not a principle. That's a preference. Anthropic pours $41M into D.C. policy, and its co-founder thanks Jia by name. Paid narrative? You tell me. So here is his tweet, in plain English: A legally bulletproof hit-piece manual. A chokehold on open source disguised as "preservation." One-way transparency. Not a shield for the nation. A moat for a valuation. And the most ironic part: Professor Jia researches law and authoritarianism. He spent his career studying how authoritarian states dress up political blacklists in "neutral" legal paperwork—only to turn around in Washington and write the how-to manual himself. What the founders—and the Court—call it. Jefferson described open source in 1813, before it had a name: "He who receives an idea from me, receives instruction himself without lessening mine; as he who lights his taper at mine, receives light without darkening me." An open model is that candle. Anthropic wants a warning label on the flame. Madison told us how freedom is actually lost. Not by "violent and sudden usurpations." By "gradual and silent encroachments of those in power." No vote. No trial. No ban. Just a label. That is what a silent encroachment looks like. And the Supreme Court has seen this exact trick: - In 1963 (Bantam Books v. Sullivan), a state commission didn't ban a single book—it just put them on a list and circulated it. The Court called it informal censorship and struck it down. - In 2024 (NRA v. Vullo), a 9–0 Court ruled the government cannot lean on third parties to punish what it cannot punish itself. So, Professor—a former clerk to two Supreme Court Justices—here is the constitutional problem you somehow missed: It isn't the wording of the label. It's the label. Punishment by list is the oldest trick of every king the founders ran from. Two chambers at One First Street, and you walked away thinking they wrote the First Amendment so the blacklist could have better grammar. Free code. Free speech. Free people. Same fight. A free country does not put warning labels on ideas. One more thing: Anthropic, @ch402 next time, spend a little more and get better "independent legal commentary."
显示更多
0
3
112
36
转发到社区