It’s surely fitting that a decision released on July 4 would set off fireworks on the Cyberlaw Podcast. The source of the drama was U.S. District Court Judge Terry Doughty’s injunction prohibiting multiple federal agencies from leaning on social media platforms to suppress speech the agencies don’t like. Megan Stifel, Paul Rosenzweig, and I could not disagree more about the decision, which seems quite justified to me, given the aggressive White House communications telling the platforms whose speech the government wanted suppressed. Paul and Megan argue that it’s not censorship, that the judge got standing law wrong, and that I ought to invite a few content moderation aficionados on for a full hour episode on the topic.   That all comes after a much less lively review of recent stories on artificial intelligence. Sultan Meghji downplays OpenAI’s claim that they’ve taken a step forward in preventing the emergence of a “misaligned”—in other words evil—superintelligence. We note what may be the first real-life “liar’s dividend” from deep faked voice. Even more interesting is the prospect that large language models will end up poisoning themselves by consuming their own waste—that is, by being trained on recent internet discourse that includes large volumes of text created by earlier models. That might stall progress in AI, Sultan suggests. But not, I predict before government regulation tries to do the same; as witness, New York City’s law requiring companies that use AI in hiring to disclose all the evidence needed to sue them for discrimination. Also vying to load large language models with rent-seeking demands are Big Content lawyers. Sultan and I try to separate the few legitimate intellectual property claims against AI from the many bogus ones.  I channel a recent New York gubernatorial candidate in opining that the rent-seeking is too damn high.  Paul dissects China’s most recent and self-defeating effort to deter the West from decoupling from Chinese supply chains. It looks as though China was so eager to punish the West that it rolled out supply chain penalties before it had the leverage to make the punishment stick. Speaking of self-defeating Chinese government policies, it looks as though the government’s two-minute hate directed at China’s fintech giants is coming to an end. Sultan walks us through the wreckage of the American cryptocurrency industry, pausing to note the executive exodus from Binance and the end of the view that cryptocurrency could be squared with U.S. regulatory authorities. Not in this administration, and maybe not in any, and outcome that will delay financial modernization here for years. I renew my promise to get Gus Coldebella on the podcast to see if he can turn the tide of negativism.  In quick hits and updates: There’s an effort afoot to amend the National Defense Authorization Act  to prevent American government agencies, and only American government agencies, from buying data available to everyone else. We are skeptical that it will pass.  The EU and the U.S. have reached a (third) transatlantic data transfer deal, and just in time for Meta, which was facing a new set of competition attacks on its data protection compliance. And Canada, which already looks ineffectual for passing a link tax that led Facebook and Google to simply drop links to Canadian media, now looks ineffectual and petty, announcing it has pulled its paltry advertising budget from Facebook. Oh, and last year’s social media villain is this year’s social media hero, at least on the left, as Meta launches Threads and threatens Twitter’s hopes for a recovery. Download 467th Episode (mp3) You can subscribe to The Cyberlaw Podcast using iTunes, Google Play, Spotify, Pocket Casts, or our RSS feed. As always, The Cyberlaw Podcast is open to feedback. Be sure to engage with @stewartbaker on Twitter. Send your questions, comments, and suggesti