Oh my debut in podcast

I worked so, so hard on this piece: It's about OpenAI bringing back erotica, what's been going on with users' mental health, and how it all relates to making AI go well.

@kashhill @nytimes @IASEAIorg @cquanze @manoelribeiro @PeterHndrsn @QVeraLiao @amyxzh @MicahCarroll @law_ai_ @claireboine @aylin_cim @yoshi_kohno @4sWeb I'm on the faculty job market this year🎓 - please connect if you're interested in my research on AI governance, human rights, and the intersection of technology, human agency, and policy. [18/18]

@kashhill @nytimes @IASEAIorg @cquanze @manoelribeiro @PeterHndrsn @QVeraLiao @amyxzh @MicahCarroll @law_ai_ @claireboine @aylin_cim @yoshi_kohno 👥I'll be presenting the core ideas from this paper tomorrow at the 4S Conference in Seattle @4sWeb. Find me at the open panel "c(AI)borg manyfesto: reverberating futures of augmentation and humanity." 8:30am-10:30am Cedar A [17/n]

@kashhill @nytimes @IASEAIorg @cquanze @manoelribeiro @PeterHndrsn @QVeraLiao @amyxzh @MicahCarroll @law_ai_ @claireboine This paper builds on our earlier work analyzing GenAI (now, very real) risk scenarios, co-authored with @aylin_cim and @yoshi_kohno in AI and Ethics. https://t.co/smjAThd8sX [16/n]

@kashhill @nytimes @IASEAIorg @cquanze @manoelribeiro @PeterHndrsn Amazing PrincetonCITP colleagues/mentors and collaborators: Steven Kelts, Dominik Stammbach, Patty Liu, Kylie Zhang, @QVeraLiao, Sofia Serrano, Nadja Schaetz, @amyxzh, @MicahCarroll. Can’t forget the Workshop on the Law-Following AI @law_ai_ Cullen O’Keefe, @claireboine [15/n]

@kashhill @nytimes @IASEAIorg @cquanze @manoelribeiro @PeterHndrsn Also, got great feedback from participants at the Freedom of Expression Scholars Conference @yaleisp (both 2024 and 2025) @MasonMarksMD, Erin Miller; Privacy Law Scholars Conference (2024-Europe and 2025-US) @rcalo, @MarijnSax, Mark Blitz, Tao Huang @HideyukiMATSUMI [15/n]

@kashhill @nytimes @IASEAIorg I learnt a lot from the "emotional attachment" workshop I co-organized with @cquanze @manoelribeiro and @PeterHndrsn https://t.co/ufMOxgdCfC [14/n]

@kashhill @nytimes This paper grew from valuable feedback from many amazing thinkers. Thanks @IASEAIorg for allowing me to present the early version to the global audience. Video clip is available: https://t.co/9jfXwF1wl0 [13/n]

@kashhill @nytimes The First Amendment should protect human expression and cognitive freedom—not algorithmic manipulation disguised as editorial discretion. Time to reclaim constitutional principles for the people they were meant to serve 👥✊ [12/n] Paper:

@kashhill @nytimes This isn't about censorship. It's about accountability. Just like we require drug companies to disclose side effects, AI companies should explain how their systems might affect human thought and emotion. Transparency serves speech, not silence. [11/n]

@kashhill @nytimes When these systems undermine users' cognitive freedom, through manipulation, opacity, or epistemic confusion, the government has an obligation to do so under First Amendment principles⚖️. [10/n]

@kashhill @nytimes However, GenAI systems serving “all” users without predetermined viewpoints are more like libraries than opinion publishers. It gets minimal First Amendment protection, just enough to support their users' speech rights, not to hide behind corporate secrecy. [9/n]

@kashhill @nytimes This obliviousness is dangerous and should be corrected. My paper argues for institutional distinctions. Not all institutions deserve the same level of constitutional protection. Churches ⛪, newspapers 📰, and universities 🏫 have clear expressive missions. [8/n]

@kashhill @nytimes This position is surprisingly well-supported by the First Amendment jurisprudence. Courts take "all speech is speech" approach regardless of who speaks, what impacts they make, how much attenuated from human thought. [7/n]

@kashhill @nytimes When regulators ask "How do these systems work? What are the risks?" companies wave the First Amendment like a magic wand. Transparency requirements become "compelled speech." Accountability becomes "censorship." [6/n]

@kashhill @nytimes Even more concerning: GenAI developers themselves don't understand their systems' intentions. In addition, users are active contributors to these relationships. This opacity complicates accountability attributions. [5/n]

(This distinction is indebted to excellent investigative articles from @kashhill @nytimes) https://t.co/hRPkJZbi9N [4/n]

𝑬𝒎𝒐𝒕𝒊𝒐𝒏𝒂𝒍 𝒉𝒂𝒓𝒎: When GenAI exploits our need for connection by creating artificial intimacy. Users form deep bonds with chatbots that feel real but aren't. The illusion of care without actual empathy. [3/n]

𝑬𝒑𝒊𝒔𝒕𝒆𝒎𝒊𝒄 𝒉𝒂𝒓𝒎: When GenAI distorts our ability to think clearly. Through bias, manipulation, and making us dependent on AI for answers, these systems can erode our critical thinking skills and lead us to lose our capacity to be credible knowers. [2/n]