Better Offline – “The World of AI Regulation” with Brian Merchant

Episode Date: October 10, 2025 | Host: Ed Zitron | Guest: Brian Merchant
A production of Cool Zone Media and iHeartPodcasts

Overview

This episode of Better Offline dives deep into the current reality of AI regulation, focusing on California’s legislative environment and the outsized influence of tech industry lobbying. Host Ed Zitron sits down with author/journalist Brian Merchant (Blood in the Machine) to explore the political theater behind recently passed—and toothless—AI legislation, the aggressive lobbying campaigns mounted by Silicon Valley, and what meaningful AI regulation might actually look like.

Key Discussion Points & Insights

1. Dissecting California’s Watered-down AI “Safety” Bill

[03:02–12:10]

• The recently signed California AI safety bill is the product of a year-long struggle by “AI Doomer”/safety advocates.
• The original version (Bill 1047) had real regulatory teeth: third-party audits, actual transparency, and data sharing mandates, but was fiercely opposed by all major AI companies, including so-called “safety-first” firms like Anthropic.
• After significant redrafting and lobbying, the bill was gutted into a “watered down” version that:
  • Requires companies to publish privacy/security policies publicly (“like a privacy policy”).
  • Implements an honor system: companies are supposed to self-report “unsafe” incidents to the state—but with no meaningful enforcement.
  • Adds whistleblower protections, though these are described as largely symbolic in the current legal climate.
  • Establishes Cal Compute, a public AI research consortium, yet essentially just creates another committee for discussion.

Memorable quote:

“It’s just an honor system to alert the state if you’ve done something wrong… Now Gavin Newsom gets to say, ‘I signed an AI bill, I’m doing my due diligence.’”
—Brian Merchant [09:48]

• Both hosts express skepticism with the system’s lack of “teeth” or any clear mechanisms for enforcement, with Ed Zitron bluntly calling it: “It will give the fake view that the AI companies are regulated so they’re able to continue doing all the actual harmful stuff.” [12:00]

2. Tech Lobbying Power and Legislative Apathy

[15:05–29:01]

• Brian details first-hand accounts from lawmakers and labor advocates about the relentless lobbying efforts by Silicon Valley.
• Even modestly ambitious regulations see their main provisions stripped via industry influence: “Bills that started out with some teeth have had most of them knocked out or have been delayed until next session.” [15:14]
• Notably, bills that tech companies actually fight hard against are seen as a “bellwether” for what might have some real effect.

Case Study: LEAD Act (AB 1064, California)

• Would require that chatbots marketed to children show they won’t encourage self-harm.

• Industry opposition is ferocious. Lobbying groups (e.g., Chamber of Progress, American Innovators Network, funded by VC firms like A16Z) are buying Facebook ads and sending lobbyists to lawmakers’ offices.

• Tech’s argument: the bill would “deprive” children of educational AI advantages.

• Brian’s take: “It’s fucking evil people…If you can’t stop your chatbot saying self-harm stuff to kids, you shouldn’t be selling it.” [18:39–19:39]

• The conversation highlights a likely technical limitation: large language models may not be reliably controllable to prevent problematic outputs.

  • Ed: “I don’t think you can actually prompt a large language model to categorically stop it doing something.” [19:34]
  • Brian: “Then you should not be selling that… to children.” [19:35]

3. Silicon Valley’s Broader Campaign Against Regulation

[26:23–28:44]

• Brian reports that this level of national, coordinated pushback and lobbying against state laws is unprecedented in 15 years of tech journalism.
• Notably, there was a recent (barely failed) campaign to outright ban all state-level AI lawmaking in the US, with a crucial vote tipped by a Republican senator safeguarding Nashville’s music industry.

Memorable moment:

“This is its own thing. This AI thing… It’s totally anti-democratic. It’s totally absurd.”
—Brian Merchant [28:45]

4. The Political Class and Reluctance to Actually Regulate

[29:01–32:13]

• Considerable blame is placed on politicians for a lack of technical understanding, ideological bias toward “not stifling innovation,” and default willingness to defer to the tech industry on their own terms.
• Even well-intentioned laws (e.g., autonomous vehicle safety requirements, union rights for gig workers) routinely get vetoed in California—sometimes with disastrous results shortly after.

Relevant exchange:

Ed (paraphrased): “Why does no one push through bills that actually address the technology—like, from people who have used it?”
Brian: “We still have this mentality—especially in the political class… It’s always been deferring to industry and then reacting.” [31:05–31:50]

5. What Should AI Regulation Look Like?

[32:13–35:52]

• Brian’s “dream” regulations and priorities include:

  • Environmental impact: Data center energy use, often worsened by loosened regulations lobbied for by AI firms.
  • Real corporate transparency and enforceable audits.
  • Strong workplace protections (e.g., against automated hiring/firing, discrimination, “robo-bossing”).
  • Fundamental labor rights for gig workers.
  • Aggressive antitrust enforcement to limit tech monopolies.
  • Much greater government skepticism of industry influence:

    “Why do you have to fucking listen to lobbyists?” —Ed Zitron [35:06]

• Both agree that, right now, Silicon Valley is “calling the shots”—and that substantive oversight is sorely lacking.

Notable Quotes and Memorable Moments

• On the hollow “honor system” of the bill:

  “No checkup policy. You need a page that people… can read. No enforcement. That’ll stop it.”
  —Brian Merchant, [06:20–06:29]

• On whistleblower protections:

  “Again, and it’s supposed to be for workers who are so concerned about catastrophic risk… not about, you know, if they see a company doing fraud or whatever. This is for catastrophic risk.”
  —Brian Merchant, [07:42–08:12]

• On the LEAD Act and tech pushback:

  “Once the Adam Rains stuff broke and OpenAI’s, you know, trying to hem and haw… we passed the Rubicon… It’s a product. It’s a software product.”
  —Brian Merchant, [18:41]

• On political deferral to industry:

  “It’s ideology. It’s just like, oh, well, like, we don’t want to stifle innovation… It’s always been this way—it’s always been deferring to industry and then reacting, right?”
  —Brian Merchant, [31:05–31:50]

• On the futility of most current regulatory efforts:

  “There’s a million things that need to happen… But right now, it’s just profoundly anti-Democratic where we’re at. Silicon Valley is just calling… They’re just hoarding the cat.”
  —Brian Merchant, [34:34–35:06]

Timestamps for Important Segments

• [03:02] – Overview of the “AI safety” bill’s journey and industry opposition
• [05:36–09:06] – Breakdown of the weakened final bill and why Brian Merchant isn’t celebrating
• [12:00–12:10] – Ed Zitron on fake views of regulation
• [15:14–17:43] – What meaningful regulations could look like; industry fear of child safety measures
• [18:39–21:56] – The real reason why companies may be unable to prevent chatbot harm, opposition logic
• [26:23–28:44] – Nationwide tech lobbying push to suppress state AI laws
• [29:01–32:13] – “Why doesn’t meaningful, user-informed regulation happen?”
• [32:13–35:52] – Dream regulations, environmental/worker protections, call for antitrust, and political reluctance
• [35:06] – Ed’s rant: “Why do you have to fucking listen to lobbyists?”

Conclusion

This episode lays bare the gap between the public conversation about “AI safety” and the practical realities of lawmaking in the shadow of Silicon Valley lobbying. Listeners get a nuanced, sometimes exasperated, but ultimately clear-eyed assessment of why so little meaningful AI regulation exists—and what, ideally, should be done to protect workers, the environment, and society overall. For anyone seeking a primer on where AI regulation really stands in 2025, and why, this is a must-listen episode.

Find Brian Merchant:

• Newsletter: bloodinthemachine.com
• Social: [@brianmerchant on most platforms]

Host: Ed Zitron

“Just because you put up a web page about ‘safety’ doesn’t mean you’re being regulated.”