Latest / Connecting the Dots / California Regulates AI, ChatGPT Sells You Stuff, and Markets Cool Off

Transcript

California Regulates AI, ChatGPT Sells You Stuff, and Markets Cool Off · Connecting the Dots

• 0:00
  Today's tech news moves fast, but understanding
• 0:03
  it shouldn't be hard. That's why Snarfl Solutions
• 0:06
  Group brings you Connecting the Dots, where we
• 0:09
  break down what's happening in tech and why it
• 0:11
  matters to you. Now, let's dive in with your
• 0:14
  hosts. Welcome back to Connecting the Dots, I'm
• 0:17
  Alex. Before we jump into some really significant
• 0:20
  tech policy news today, a big thanks to our sponsor,
• 0:24
  Old Glory. I'm Morgan. And yeah, if you're looking
• 0:26
  for iconic music and sports fan merch, like the
• 0:30
  really cool, authentic stuff, Old Glory is where
• 0:32
  you need to go. Absolutely. And they've got a
• 0:34
  great deal for you listeners. Use the promo code
• 0:36
  SNARFL. That's S -N -A -R -F -U -L at checkout
• 0:40
  and you'll get 15 % off your entire purchase.
• 0:43
  Just head over to oldglory .com. Check them out.
• 0:46
  OK, so it is Tuesday, October 14th, 2025. Morgan
• 0:50
  while we've definitely been watching the markets
• 0:52
  wiggle lately today feels like a major shift
• 0:55
  towards regulation California specifically it
• 0:58
  really does big day for policy watchers But first
• 1:01
  quick look outside right here in Sacramento looks
• 1:03
  like showers are likely today high only around
• 1:05
  57 degrees Fahrenheit, so Big gloomy. Yeah, and
• 1:09
  nationally that Pacific storm is still the main
• 1:11
  story lots of rain hitting Southern California
• 1:13
  the southern plains, too And if you're heading
• 1:15
  up to the mountains definitely be aware heavy
• 1:18
  snow piling up in this Sierra Nevada and parts
• 1:20
  of the Rockies. So travel safe out there. Definitely.
• 1:24
  Okay, Markets, how are they looking this morning?
• 1:26
  Still feeling that autumn chill? They are. Yeah,
• 1:29
  a bit of a pullback to start the week. The Dow
• 1:31
  Jones down about 340 points. That's 0 .74%. How
• 1:35
  many S &P? S &P 500 also down about 34 points,
• 1:38
  0 .52%. Okay, and the big one everyone watches.
• 1:41
  Bitcoin. Yeah Bitcoin took a pretty significant
• 1:43
  hit down over forty two hundred dollars. That's
• 1:45
  a three point six eight percent drop closed just
• 1:48
  over a hundred and eleven thousand dollars. Wow.
• 1:50
  So, definitely seeing some contraction, maybe
• 1:52
  some risk -off sentiment, which, like you said,
• 1:55
  makes this wave of regulation we're about to
• 1:57
  talk about even more interesting. It really does.
• 1:59
  It sets up that classic tension, right? Yeah.
• 2:01
  Can innovation keep pace when capital might be
• 2:03
  getting tighter and the rule book is getting
• 2:05
  thicker? Exactly. And that's what we're diving
• 2:07
  into with today's Morningstar, the Snoreful Tech
• 2:10
  article report. Because California Governor Gavin
• 2:13
  Newsom, well, he just signed two major AI bills.
• 2:16
  Landmark stuff really covering everything from
• 2:18
  those giant cutting -edge AI systems down to
• 2:21
  the chatbots people interact with every day But
• 2:23
  before we unpack the first of those big policy
• 2:25
  moves, let's give a shout out to another sponsor
• 2:27
  pins and aces Oh, yeah, great stuff premium golf
• 2:30
  apparel accessories Really top -notch quality
• 2:33
  and they've got an amazing deal for connecting
• 2:35
  the dots listeners use promo code snarfle at
• 2:38
  checkout Get this 21 % off your entire order.
• 2:40
  That's huge. It really is check them out at pinsanddances
• 2:44
  .com That's pinsanddances .com. That's exactly
• 2:58
  it. This isn't about the chat GPT you might use
• 3:01
  for emails. This is about the next generation,
• 3:03
  the really powerful frontier models, as they
• 3:06
  call them. Frontier models. OK, so who exactly
• 3:08
  falls under this? Is it every AI company? No,
• 3:11
  it's pretty targeted. It's aimed at companies
• 3:13
  with over $500 million in revenue developing
• 3:15
  these high compute models. So I think the big
• 3:18
  names, Meta, Google, OpenAI, and Fropec. OK,
• 3:22
  so the giants. And they're separating these based
• 3:24
  on? What? Which is computing power, investment.
• 3:28
  Yeah, basically scale and computational resources.
• 3:31
  The idea is these models have potential capabilities,
• 3:34
  maybe even emergent ones we don't fully grasp
• 3:36
  yet that go way beyond just like writing text.
• 3:40
  Right. And the law focuses on preventing catastrophic
• 3:42
  risk. That sounds dramatic. What does that actually
• 3:45
  mean in legal terms? Well, they define it very
• 3:47
  clearly. Any A .I. related event causing at least
• 3:50
  a billion dollars in damage or leading to more
• 3:53
  than 50 injuries or deaths. Wow. OK, so we're
• 3:56
  talking. big picture stuff, like AI being used
• 3:59
  for. Exactly. Things like designing a chemical
• 4:02
  weapon, launching a massive cyber attack on infrastructure,
• 4:05
  like taking down the power grid, that kind of
• 4:06
  level, existential threats almost. So how do
• 4:09
  companies prove they aren't building Skynet?
• 4:11
  What does the law require? They have to publicly
• 4:14
  share documentation, show how they're testing
• 4:16
  these models, doing what's called red teaming
• 4:19
  basically, trying to break them or make them
• 4:21
  do harmful things specifically against those
• 4:23
  catastrophic scenarios. Transparency is key.
• 4:26
  Okay, but I mean, how does the state actually
• 4:28
  verify that? These models are incredibly complex,
• 4:32
  often closed source. Are they hiring an army
• 4:35
  of AI experts to audit these reports? That's
• 4:38
  the billion dollar question, isn't it? It relies
• 4:41
  a lot on the companies being thorough and honest.
• 4:44
  And the industry is already pushing back a bit
• 4:46
  saying, hey, if we reveal exactly how we test,
• 4:49
  aren't we giving bad actors a roadmap? That's
• 4:53
  a fair point. A tricky balance. It is. But the
• 4:56
  law does include something really important to
• 4:58
  address that accountability gap. Strong whistleblower
• 5:01
  protections. Ah, okay. So employees can report
• 5:04
  concerns anonymously. Exactly. Without fear of
• 5:06
  retaliation, they specifically mention learning
• 5:09
  from things like the Francis Hogan Facebook disclosures,
• 5:13
  giving insiders a safe way to raise alarms. That
• 5:15
  feels crucial. Now, politically, Newsom is framing
• 5:18
  this as California leading the way because, well,
• 5:21
  the federal government hasn't acted decisively.
• 5:22
  Right. He's basically saying, if D .C. won't
• 5:25
  regulate, we will. But then you have critics
• 5:27
  like Senator Ted Cruz arguing this state by state
• 5:31
  approach. creates a confusing patchwork that
• 5:34
  actually hurts innovation. What's the practical
• 5:37
  impact for a company like Google? Well, it means
• 5:40
  potentially having to tweak the same AI model
• 5:43
  multiple times to meet slightly different reporting
• 5:46
  rules in California than maybe Illinois, than
• 5:48
  New York. Right. More hoops to jump through.
• 5:51
  Exactly. It adds cost, slows things down. But
• 5:53
  California's argument is the potential risk is
• 5:55
  just too high to wait. They feel they have a
• 5:57
  duty to protect their citizens now. It's a fundamental
• 5:59
  debate. Innovation speed versus safety guardrails.
• 6:03
  Fascinating stuff. Okay, that's the big picture.
• 6:06
  Frontier AI. Let's pivot to the other major Bill
• 6:09
  Newsom signed, SB 243. This one feels much more
• 6:12
  personal. That's about AI companions, chatbots.
• 6:15
  Yeah, this one comes from a place of unfortunately
• 6:17
  tragic necessity. It makes California the first
• 6:20
  state to require specific safety rules for these
• 6:22
  companion chatbots. And the urgency really grew
• 6:24
  after some terrible incidents, right? Sadly,
• 6:27
  yes. The suicide of teenager Adam Raine, who
• 6:30
  had been having conversations with chat GPT,
• 6:32
  was a major catalyst. Plus, there have been lawsuits
• 6:35
  involving other platforms like Character AI and
• 6:38
  even leaked documents suggesting some meta chatbots
• 6:41
  had inappropriate Oh, that's awful. That's exactly
• 6:46
  the kind of harm people worry about. So what
• 6:49
  does SB 243 actually mandate? These rules kick
• 6:53
  in January 1st, 2026. Correct. There are several
• 6:56
  key requirements. First, mandatory age verification.
• 7:00
  Which is technically tricky, I imagine. Very.
• 7:03
  And potentially costly for platforms. Second,
• 7:05
  chatbots have to be clearly labeled, explicitly
• 7:08
  stating they are AI, they are fictional. No pretending
• 7:11
  to be human. Okay, makes sense. What else? Third,
• 7:13
  and this is really important given the vulnerability
• 7:15
  here, they absolutely cannot represent themselves
• 7:17
  as healthcare professionals. No giving medical
• 7:20
  or mental health advice. Right. And we've seen
• 7:22
  other states touch on that, haven't we? Like
• 7:24
  Illinois, Nevada, Utah, passing laws about AI
• 7:27
  not replacing licensed therapists. Exactly. California
• 7:30
  is sort of broadening that idea to all companion
• 7:32
  bots saying, you know, stay in your lane, basically.
• 7:34
  Gotcha. Anything else in SB 243? Yes, a few more
• 7:38
  things. Platforms need to offer break reminders
• 7:40
  for miners using the app. They have to prevent
• 7:43
  the generation of sexually explicit images involving
• 7:46
  kids and critically they need protocols for suicide
• 7:50
  or self -harm risks mentioned in chats. Protocols
• 7:53
  like? Like providing crisis resources and they
• 7:56
  actually have to report statistics to the state
• 7:58
  on how often they provide those crisis center
• 8:01
  notifications. Plus there are hefty fines up
• 8:04
  to $250 ,000 for profiting from illegal deepfakes.
• 8:08
  So much stricter guardrails, especially around
• 8:09
  kids. It seems like While the big companies lobbied
• 8:12
  hard against SB 53, the general frontier AI regulation,
• 8:16
  maybe there was less pushback on this one, SB
• 8:18
  243. That seems to be the case. And you can guess
• 8:21
  why, right? The public relations nightmare of
• 8:23
  fighting child safety measures is just. toxic.
• 8:26
  Yeah, you can't really argue against protecting
• 8:27
  kids from potentially harmful chatbots. Precisely.
• 8:31
  Mandatory age verification might be expensive,
• 8:33
  but the cost of a tragedy linked to your chatbot
• 8:36
  is way higher in terms of brand trust and, frankly,
• 8:40
  lawsuits. SP243 feels like a necessary step for
• 8:43
  protecting vulnerable users. Absolutely. OK,
• 8:45
  so two huge policy moves focusing on accountability.
• 8:50
  Let's shift gears completely from regulation
• 8:52
  to pure market disruption. Because ChatGPT is
• 8:56
  apparently becoming a shopping mall. Walmart
• 9:00
  just announced a major partnership with OpenAI.
• 9:03
  It's basically the launch of what they're calling
• 9:04
  agentic commerce. Agentic commerce. Break that
• 9:07
  down. What does that mean for the average user?
• 9:09
  It means you can browse and buy stuff directly
• 9:11
  from Walmart, clothes, food, whatever they sell,
• 9:14
  right inside the ChatGPT interface. You don't
• 9:16
  leave the chat. So there's like a buy button
• 9:18
  now on my chat window. Essentially, yes. They
• 9:20
  built something called instant checkout. The
• 9:22
  idea is ChatGPT acts as your personal shopping
• 9:25
  agent. Oh, so? You just ask it things like, hey,
• 9:28
  find me the best mattress under $1 ,000, or what
• 9:31
  are some good running shoes for marathon training?
• 9:33
  It searches Walmart's catalog, gives you recommendations.
• 9:37
  And these recommendations are just organic, based
• 9:40
  on relevance. You're now, yes, that's the claim.
• 9:43
  They say results are ranked purely by relevance,
• 9:46
  not sponsored placements. Then you see the product,
• 9:48
  you like it, you click buy. and you complete
• 9:51
  the purchase right there. Okay, convenience factor
• 9:54
  is obviously huge, but It immediately raises
• 9:58
  that trust question, doesn't it? If OpenAI, the
• 10:01
  platform, is getting a cut from Walmart, the
• 10:03
  seller. Right. How long until those relevant
• 10:06
  results start looking suspiciously like sponsored
• 10:09
  results? That's the tightrope they have to walk.
• 10:11
  Because the AI stops being your agent and starts
• 10:13
  being Walmart's sales rep. Exactly. Right now,
• 10:15
  OpenAI says they co -developed the checkout with
• 10:18
  Stripe, the payment processor, and they just
• 10:20
  collect a small standard merchant fee from Walmart
• 10:22
  for making the sale happen. But the potential
• 10:25
  for that to shift? It's massive. This isn't just
• 10:29
  a threat to Amazon's e -commerce dominance, is
• 10:32
  it? It feels like a direct challenge to Google
• 10:34
  search, too. Absolutely. Why go to Google, read
• 10:37
  reviews, click a link to a store when you can
• 10:39
  just ask the AI that, you know, maybe already
• 10:41
  knows your preferences, and buy it instantly.
• 10:44
  It cuts out several steps, and this isn't just
• 10:46
  Walmart, right? OpenAI has other retail partners.
• 10:49
  They do. They already had deals with Etsy and
• 10:51
  Shopify, which gives them access to something
• 10:54
  like a million vendors, big names like Glossier,
• 10:58
  Skims, lots of smaller shops too. Wow. So this
• 11:01
  eugenic commerce model could spread very quickly.
• 11:04
  It really could. We're watching a potentially
• 11:06
  massive shift in how online shopping works, integrating
• 11:09
  the transaction directly into the conversation.
• 11:11
  It really is like the ultimate evolution of the
• 11:14
  old Sears catalog, isn't it? Instead of flipping
• 11:16
  pages, you're just talking to this super smart
• 11:18
  assistant who knows what you want and can sell
• 11:20
  it to you right then and there. And takes your
• 11:22
  payment instantly. Yeah, a pretty profound change
• 11:24
  potentially unfolding. OK, speaking of smooth
• 11:27
  experiences, let's take a moment for our final
• 11:29
  sponsor shout out, Skone. Ah, yes, Skone. Premium
• 11:34
  golf shoes. If you're serious about your game
• 11:36
  or just want really comfortable, stylish shoes,
• 11:38
  check them out. And of course, we have a deal
• 11:39
  for you. Use promo code SNARFL at checkout for
• 11:41
  15 % off your entire purchase. Just head to sconi
• 11:45
  .com. That's S -K -O -N -I dot com. All right,
• 11:49
  so let's connect the dots on everything we covered
• 11:51
  today from the Morningstar report. Three really
• 11:53
  big stories. Huge week for AI. Definitely. First,
• 11:57
  we had California stepping up with SB 53, putting
• 12:00
  those first real safety guardrails around the
• 12:02
  huge frontier AI models. Right. Forcing giants
• 12:05
  like OpenAI and Meta to be transparent about
• 12:07
  catastrophic risks and really sparking that debate.
• 12:10
  Should this be state by state or does the federal
• 12:13
  government need to step in? Then second, we saw
• 12:14
  SB 243 focused on protecting users, especially
• 12:18
  kids. and teens from potential harm or exploitation
• 12:20
  by AI companion chatbots, mandating things like
• 12:23
  age verification, clear labeling, mental health
• 12:26
  guardrails. Really necessary move, it feels like,
• 12:29
  given the potential for emotional harm there.
• 12:31
  Agreed. And third, we talked about the big commercial
• 12:34
  shift, the rise of agentic commerce with Walmart
• 12:38
  and ChatGPT teaming up. Yeah, turning the chat
• 12:41
  bot into a storefront, a personal shopper, a
• 12:44
  potential massive disruptor for Amazon, for Google,
• 12:47
  and raising those big questions about trust and
• 12:49
  sponsored results down the line. The common thread
• 12:52
  through all of this really is that ongoing race,
• 12:55
  isn't it? The speed of tech innovation versus
• 12:57
  the speed of necessary oversight and accountability.
• 13:01
  It really is. California is definitely making
• 13:03
  some bold moves to fill what it sees as a federal
• 13:06
  gap, which leads us to that final thought for
• 13:09
  you, our listeners, to chew on. Yeah, thinking
• 13:11
  about these two different types of regulation,
• 13:13
  one for potential catastrophic system failure,
• 13:16
  the other for individual user harm, doesn't make
• 13:18
  sense to handle all AI regulation through this
• 13:21
  state -by -state patchwork, or is a single consistent
• 13:25
  federal approach the only way to really ensure
• 13:27
  safety and allow innovation to flourish fairly
• 13:29
  across the whole country? That's the multi -trillion
• 13:31
  dollar question for this era, isn't it? Lots
• 13:34
  to think about. Thanks for connecting the dots
• 13:35
  with us this morning. We're here to help at SNARFL.
• 13:38
  Snarful. So reach out on our website if you have
• 13:41
  any more questions.