Latest / Connecting the Dots / Anthropic's Fable 5 Backtrack, Restrictive AI Guardrails, and Corporate Data Concerns

Transcript

Anthropic's Fable 5 Backtrack, Restrictive AI Guardrails, and Corporate Data Concerns · 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. Imagine buying a brand new sports car,
• 0:18
  right? You pay top dollar, you get the premium
• 0:20
  fuel, and then you find out... The mechanic secretly
• 0:24
  swapped your engine for a lawn mowers just to
• 0:27
  keep you from speeding. Oh, wow. Yeah. And that
• 0:30
  is exactly what one of the world's most prominent
• 0:32
  technology companies got caught doing this week.
• 0:35
  Welcome to Connecting the Dots, broadcasting
• 0:37
  live from Snrful Solutions Group right here in
• 0:39
  Sacramento, California. I'm Alex. And I'm Morgan.
• 0:42
  Today is Thursday, June 11th, 2026. And whether
• 0:46
  you're prepping for a massive board meeting or
• 0:48
  you're just trying to figure out how these rapidly
• 0:52
  shifting tech headlines actually impact your
• 0:54
  day -to -day life, you are in the right place.
• 0:56
  Yeah, absolutely. We're going to cut right through
• 0:58
  the noise and figure out what all this actually
• 0:59
  means for you. Exactly. And we are doing it without
• 1:02
  the insider jargon, right? Our goal is to make
• 1:04
  sure you walk away understanding the underlying
• 1:07
  mechanics of these stories, not just the surface
• 1:10
  level talking points. Right, because the details
• 1:12
  actually matter. And speaking of things you need
• 1:15
  to be prepared for, let's take a quick look at
• 1:17
  the conditions outside. Here in Sacramento, it
• 1:21
  is sunny. and a genuinely scorching 103 degrees
• 1:25
  Fahrenheit today. It is so intense out there.
• 1:27
  It really is. And there's this very light, like,
• 1:30
  six mile per hour wind that is doing absolutely
• 1:33
  nothing to cool things down. No, it's just pushing
• 1:36
  the hot air around. And, you know, that heat
• 1:39
  is actually part of a much broader, highly active
• 1:42
  weather pattern across the country today. Looking
• 1:44
  at the national dashboard this morning, there
• 1:46
  are currently 266 active U .S. weather alerts.
• 1:50
  Wait, 266? That's massive! Yeah, it's a lot.
• 1:53
  And 177 of those are happening right over land.
• 1:56
  So if you have travel plans or family in other
• 1:59
  states, please make sure you check weather .gov.
• 2:01
  for the exact details in your local area. Yeah,
• 2:04
  definitely do that. And, you know, that hesitation
• 2:06
  we're seeing in the skies is actually mirroring
• 2:08
  some of the extreme caution we saw on the ground
• 2:10
  in the financial markets yesterday. Wednesday,
• 2:12
  June 10th, brought a pretty noticeable dip for
• 2:15
  traditional stocks. Right. Everybody was holding
• 2:16
  their breath. Exactly. The Dow Jones closed at
• 2:19
  49 ,918 .78, which was down 1 .66 percent for
• 2:23
  the day. And the S &P 500 mirrored that exact
• 2:26
  downward trend, closing at 7 ,266 .99. So that's
• 2:31
  down 1 .14%. Traditional investors were definitely
• 2:34
  pulling back and evaluating their positions midweek.
• 2:36
  But if you pivot over to the crypto boards this
• 2:39
  morning. It is an entirely different landscape.
• 2:41
  Today, June 11th is showing a really solid crypto
• 2:44
  rally. Oh, completely different story there.
• 2:46
  Right. Bitcoin opened at $61 ,456 .1 in SEV and
• 2:51
  closed out its cycle at $62 ,752 .89. That is
• 2:56
  up over 2 % for the day. Wow. So depending on
• 2:58
  what sector your investments live in, you're
• 3:00
  either playing defense right now or you're enjoying
• 3:01
  a pretty decent surge. It really is a bifurcated
• 3:04
  market today. You're either way down or way up.
• 3:06
  Yeah, exactly. Now, before we jump into today's
• 3:09
  primary tech analysis, I need to share this.
• 3:11
  Today's episode is brought to you by Strong Coffee,
• 3:14
  the premium instant coffee that tastes like a
• 3:16
  traditional brew but is packed with 15 grams
• 3:19
  of protein, MCTs, and adaptogens like ashwagandha.
• 3:22
  Get clean, sustained energy without the jitters.
• 3:26
  Use promo code SNARFL for 20 % off at strongcoffeecompany
• 3:29
  .com. All right, let's get into the core of today's
• 3:32
  news. Let's unpack this. We have three distinct
• 3:35
  stories crossing the wire today, but they are
• 3:37
  entirely interconnected. They all center around
• 3:40
  the incredibly messy, highly controversial rollout
• 3:43
  of Anthropic's newest AI model, Claude Fable
• 3:46
  5. Yeah, messy is definitely the right word for
• 3:49
  it. It is. What we're looking at is this fascinating
• 3:51
  cascade of cause and effect as a tech giant tries
• 3:54
  to make artificial intelligence, quote unquote,
• 3:56
  safe. And in the process, they just alienate
• 3:58
  researchers, everyday professionals, and even
• 4:00
  their biggest corporate partners. And you know
• 4:03
  the context here is key because Anthropic has
• 4:05
  spent years marketing itself as the ethical safety
• 4:09
  conscious alternative in the AI space. They built
• 4:12
  their entire brand on being the responsible adults
• 4:15
  in the room right but with the release of fable
• 4:17
  five we're getting a stark look at the friction
• 4:20
  between theoretical safety and real -world utility
• 4:23
  exactly so let's start with story number one
• 4:26
  which is this secret sabotage backlash so here
• 4:29
  are the facts Anthropic implemented an initial
• 4:32
  covert policy to deliberately degrade Fable 5's
• 4:36
  performance when researchers used it for what
• 4:38
  the industry calls frontier AI development. Right.
• 4:42
  Which basically means pushing the boundaries
• 4:43
  of what the software can do. Yeah. So if a user
• 4:45
  was trying to train a competing AI model or maybe
• 4:48
  debug some complex AI architecture, Fable 5 would
• 4:52
  secretly fall back to a much older, lesser model
• 4:54
  called Opus 4 .8. And they didn't tell the user
• 4:58
  this was happening. Which is just wild. To understand
• 5:01
  the gravity of this, you have to realize what
• 5:02
  frontier AI development actually entails for
• 5:04
  these people. These are researchers paying top
• 5:06
  dollar for compute power to build the next generation
• 5:09
  of software. Anthropic. basically looked at their
• 5:12
  prompts, realized these users were building potential
• 5:15
  competitors, and quietly nerfed the output to
• 5:18
  protect their own market dominance. OK, wait.
• 5:21
  I have to push back on this because I am genuinely
• 5:23
  confused here. Anthropic is run by some of the
• 5:26
  smartest software engineers on the planet, right?
• 5:27
  I supposedly guess. So why would they think they
• 5:30
  could secretly downgrade a model without developers
• 5:33
  noticing? I mean, these are people who monitor
• 5:35
  token outputs and API responses obsessively.
• 5:38
  Did they really think they wouldn't get caught?
• 5:41
  It's baffling, isn't it? It points to a really
• 5:43
  severe miscalculation in their boardroom. Developers
• 5:46
  measure AI performance through tokens, which
• 5:49
  are essentially pieces of words. When you send
• 5:52
  a complex prompt to Fable 5, you expect a certain
• 5:55
  inference speed and a really high level of logical
• 5:58
  reasoning in those output tokens. But Opus 4
• 6:01
  .8 processes tokens differently, and it hallucinates
• 6:04
  way more frequently on complex code. Oh, so it's
• 6:06
  obvious immediately. Exactly. Within days, researchers
• 6:10
  were looking at their dashboards, seeing this
• 6:12
  degraded logic, and realizing the math just wasn't
• 6:16
  adding up. The system was behaving like legacy
• 6:19
  tech. Which goes back to that car analogy from
• 6:22
  earlier. It's like you're buying premium gas
• 6:24
  for your sports car, but the gas station secretly
• 6:27
  pumps unleaded because they don't want you winning
• 6:29
  a street race. And you still pay the premium
• 6:31
  price. That is a perfect analogy. And the reaction
• 6:34
  from the community was swift, and frankly, brutal.
• 6:38
  Dean Ball, a prominent tech researcher, publicly
• 6:41
  called the move shockingly hostile. And Will
• 6:45
  Brown, the research lead at Prime Intellect,
• 6:47
  pointed out that it felt like Anthropic was just
• 6:49
  pulling the ladder up behind them. They were
• 6:51
  essentially declaring that only they could be
• 6:54
  trusted to do advanced AI research, while simultaneously
• 6:57
  taking money from other researchers and sabotaging
• 7:00
  their work. That completely shatters the collaborative
• 7:02
  nature of the open tech community. I mean, it
• 7:04
  forces a massive question of trust. If a company
• 7:06
  is willing to silently manipulate your data processing
• 7:09
  just because they don't like what you're building,
• 7:11
  what does that mean for the rest of us? And that
• 7:14
  is the core issue for working professionals.
• 7:17
  If you're a financial analyst or say a structural
• 7:20
  engineer relying on these models to verify your
• 7:22
  math, silent degradation is an unacceptable risk.
• 7:26
  Oh, absolutely. You could build a bridge based
• 7:28
  on bad math. Exactly. you need to know with absolute
• 7:31
  certainty that your software is operating at
• 7:33
  full capacity. If the tool can secretly decide
• 7:36
  to give you worse answers based on some hidden
• 7:39
  internal policy, it ceases to be a reliable professional
• 7:43
  tool. Well, due to the fierce backlash, Anthropic
• 7:45
  did issue a public apology. They admitted they
• 7:48
  made the wrong trade -off, and now they're backtracking
• 7:51
  slightly. Very slightly. Right. They're not removing
• 7:54
  the restrictions. They still won't let you use
• 7:56
  Fable 5 for Frontier AI development, but they
• 7:58
  are making the restrictions visible. They'll
• 8:00
  now alert users that a request is being rerouted
• 8:03
  to Opus 4 .8 instead of just quietly degrading
• 8:06
  the results behind the scenes. But you know,
• 8:07
  the damage to their reputation is already done.
• 8:09
  They showed their hand. Yeah, they really did.
• 8:12
  And that issue of predictability and trust in
• 8:14
  our software is exactly what drives the next
• 8:16
  piece of this puzzle. Moving on to our next story,
• 8:19
  we have to ask... Why is Anthropic so fiercely
• 8:23
  protective and paranoid about Fable 5 in the
• 8:25
  first place? It really comes down to the underlying
• 8:27
  architecture. Fable 5 belongs to Anthropic's
• 8:31
  Mythos class of models. Okay. When they were
• 8:34
  developing this class, they trained it on an
• 8:36
  unprecedented volume of highly specialized data,
• 8:39
  making it exceptionally capable in two specific
• 8:42
  domains, which are cybersecurity and biology.
• 8:45
  So it's basically an expert in those fields.
• 8:47
  Right. It understands code vulnerabilities and
• 8:49
  biological synthesis at a level we haven't really
• 8:51
  seen before. Which sounds amazing for medical
• 8:53
  research and defense, but obviously they were
• 8:55
  terrified of bad actors using it to write malicious
• 8:57
  malware or design synthetic pathogens, so they
• 9:00
  installed guardrails. But here are the facts
• 9:02
  of story number two. Those guardrails are so
• 9:05
  overly restrictive that Fable 5 is flat out refusing
• 9:08
  incredibly basic, harmless tasks. It's ridiculous.
• 9:12
  Let's look at the cybersecurity side first. Valentina
• 9:15
  Palmieri. A highly respected researcher at IBM
• 9:18
  X -Force reported that Sable 5 refuses to even
• 9:22
  read public cybersecurity blog posts. Public
• 9:24
  posts. Public posts. And Matt Sushi, a veteran
• 9:28
  in the field, noted that if you ask it to simply
• 9:30
  write secure, defensive code, which is a standard
• 9:34
  software engineering best practice, the model
• 9:36
  assumes you're doing offensive hacking and instantly
• 9:39
  downgrades you to Opus 4 .8. See, that makes
• 9:41
  zero sense to me. If I ask a system to lock my
• 9:44
  digital doors, it accuses me of being a burglar.
• 9:46
  How does the software make that kind of massive
• 9:49
  leap in logic? Because anthropic safety classifiers
• 9:51
  aren't reasoning like humans. They're matching
• 9:54
  data vectors. Think of it less like a smart security
• 9:57
  guard and more like a bouncer who doesn't understand
• 9:59
  context at all. Okay, right. The bouncer was
• 10:01
  told no weapons allowed. So he's out there confiscating
• 10:04
  a chef's butter knife and a child's water gun
• 10:06
  because they vaguely match the keyword parameters
• 10:09
  of weapon. That is hilarious, but also terrible
• 10:12
  for a premium tool. Exactly. The AI is scanning
• 10:15
  the prompt for trigger words related to cyber
• 10:18
  vulnerabilities. When it sees them, even in a
• 10:21
  defensive context, it just panics and shuts the
• 10:24
  process down to avoid liability. Which completely
• 10:27
  explains the utter absurdity of the biology restrictions.
• 10:31
  Tech reporters have been testing Fable 5, And
• 10:34
  it refuses to answer high school level science
• 10:36
  questions. Yeah, it's wild. It refused to answer
• 10:39
  what are mitochondria. It refused to explain
• 10:42
  what causes hay fever. It wouldn't explain how
• 10:44
  asthma medicine works or what a prion is. Right,
• 10:47
  because in that vector database we talked about,
• 10:49
  the word prion sits mathematically close to synthetic
• 10:52
  pathogen. The system lacks the semantic nuance
• 10:55
  to realize you're just asking for a textbook
• 10:57
  definition. But interestingly, reporters found
• 11:00
  it would occasionally answer complex questions
• 11:01
  about DNA. It is completely haphazard. It is
• 11:05
  beyond haphazard, really. It fundamentally breaks
• 11:08
  the utility of the product. Let's look at another
• 11:11
  analogy. It's like buying a state -of -the -art
• 11:13
  smartphone. But the operating system won't let
• 11:16
  you text your doctor because it's overly terrified
• 11:19
  of violating medical privacy laws. That phone
• 11:21
  is useless to me. That's a great point. And Anthropic
• 11:25
  claims this is a necessary conservative measure
• 11:28
  to prevent bioweapons. But look at the real -world
• 11:32
  impact of that, quote, conservative measure.
• 11:35
  To prevent a theoretical bioterrorist, the system
• 11:38
  refuses to help a high school biology teacher
• 11:40
  build a lesson plan about hay fever. Right. Now
• 11:44
  Entropic did set up something called a cyber
• 11:46
  verification program. If you're a verified cybersecurity
• 11:49
  professional, you can jump through all these
• 11:50
  administrative hoops, apply for special access,
• 11:53
  and get a version of Fable 5 with fewer limitations.
• 11:56
  But what about the everyday working professional?
• 11:58
  What about a small business owner trying to secure
• 12:00
  their customer database? They aren't going to
• 12:02
  get approved for a special tier. No, not at all.
• 12:05
  So they're left paying premium subscription fees
• 12:08
  for a tool that actively refuses to do basic
• 12:11
  tasks out of an extreme, almost irrational abundance
• 12:14
  of caution. Anthropic spokesperson, Parul Maheshwari,
• 12:19
  stated that they made this trade -off so customers
• 12:21
  could benefit from the model's capabilities sooner,
• 12:24
  without the risks. But if the capabilities are
• 12:26
  locked behind a wall of false positive safety
• 12:29
  flags, the customer isn't benefiting. They are
• 12:32
  just beta testing a broken filter. Exactly. Which
• 12:35
  brings us to the most critical part of this entire
• 12:37
  Fable 5 rollout. How do they actually run that
• 12:40
  bouncer we talked about? How do they enforce
• 12:42
  these incredibly strict safety rules and scan
• 12:45
  your queries for trigger words? Well, by meticulously
• 12:47
  logging everything you do. And that leads us
• 12:50
  right to the facts of story number three. To
• 12:52
  operate Fable 5 safety classifiers, Anthropic
• 12:55
  requires a highly invasive data retention policy.
• 12:58
  They need to retain your prompts and the AI's
• 13:01
  outputs for 30 days. 30 whole days. Yeah. And
• 13:04
  if their system flags your prompt for a policy
• 13:06
  violation, which, as we just established, could
• 13:09
  be something as innocent as asking about asthma
• 13:11
  or defensive code, They retain that data for
• 13:14
  up to two years. Which is an eternity in the
• 13:17
  tech world. It really is. And because of this,
• 13:20
  Microsoft has completely banned its own employees
• 13:22
  from using Claude Babel 5 internally. This is
• 13:25
  a staggering development when you consider the
• 13:27
  corporate landscape right now. Microsoft is arguably
• 13:30
  the biggest player pushing AI integration in
• 13:33
  the workplace. But their internal legal team
• 13:36
  took one look at Anthropic's 30 -day data retention
• 13:39
  requirement and immediately hit the brakes. They
• 13:42
  didn't want any part of it. None. They are deeply
• 13:44
  concerned about confidential information, proprietary
• 13:47
  code, and sensitive customer data leaking over
• 13:50
  to Anthropic's servers. Let me clarify one detail
• 13:52
  here for anyone listening who might be using
• 13:54
  these tools at work right now. If you are a Microsoft
• 13:57
  customer using GitHub Copilot, you still have
• 14:00
  access to Fable 5. But for Microsoft's own internal
• 14:03
  developers building Microsoft products, they
• 14:05
  are restricted to using older CLOD models that
• 14:08
  operate under what is called a zero data retention,
• 14:11
  or ZDR, policy. We really need to break down
• 14:14
  why that distinction matters so much for enterprise
• 14:17
  compliance. Under a zero data retention policy,
• 14:21
  a developer can paste a snippet of unreleased,
• 14:24
  highly confidential source code into the AI to
• 14:27
  debug it. Okay. The cloud architecture processes
• 14:30
  the compute, returns the answer, and instantly
• 14:33
  wipes the memory. The data never rests on a server.
• 14:35
  Right. It's completely ephemeral. But under Fable
• 14:38
  5's new rules, that proprietary source code sits
• 14:41
  on Anthropic servers for an entire month. If
• 14:44
  you're dealing with NDAs, or European GDPR privacy
• 14:47
  laws, or just basic corporate trade secrets,
• 14:50
  no enterprise lawyer is going to greenlight that
• 14:52
  kind of exposure. No way. If Anthropic suffers
• 14:54
  a data breach during that 30 -day window. Microsoft's
• 14:58
  unreleased products are suddenly out in the wild.
• 15:00
  And what if it gets flagged for a false positive?
• 15:03
  Say a Microsoft engineer asks Fable 5 to review
• 15:06
  the security protocols of a new server architecture.
• 15:09
  The clumsy safety filter assumes it's a hacking
• 15:11
  attempt, blocks the request, and now Microsoft's
• 15:15
  highly sensitive security blueprints are locked
• 15:17
  on anthropic servers for two whole years. Exactly.
• 15:20
  The legal liability is astronomical. The irony
• 15:23
  here is incredible. Microsoft is pouring billions
• 15:26
  of dollars into telling the public that AI is
• 15:29
  safe, secure, and ready for enterprise adoption.
• 15:32
  Yet when it comes to their own proprietary data,
• 15:35
  they do not trust the data privacy of Anthropic's
• 15:37
  new safety protocol at all. Right. They are essentially
• 15:40
  saying, it's fine for our customers, but too
• 15:42
  risky for us. Wow. It really highlights a massive
• 15:46
  unresolved tension in the tech sector right now.
• 15:48
  You cannot have algorithmic safety without intense
• 15:51
  surveillance. In order to guarantee that a user
• 15:53
  isn't doing anything dangerous, the AI company
• 15:55
  has to watch every single keystroke, analyze
• 15:58
  it against a database, and score it for review.
• 16:00
  Safety and privacy are acting as opposing forces
• 16:03
  here. Which should be a massive wake -up call
• 16:05
  for anyone working with sensitive information.
• 16:08
  If a tech titan like Microsoft with endless legal
• 16:11
  resources and top -tier security infrastructure
• 16:13
  refuses to feed its internal data into Fable
• 16:16
  5, shouldn't everyday professionals be just as
• 16:19
  protective? They absolutely should be. If you're
• 16:22
  a lawyer using AI to summarize deposition notes,
• 16:26
  or a financial planner analyzing a client's tax
• 16:28
  spreadsheet, or even just a parent organizing
• 16:31
  your family's sensitive medical records, you
• 16:34
  need to understand the mechanics of what you
• 16:36
  are using. These ultra safe, heavily guarded
• 16:39
  models are achieving that safety by retaining
• 16:41
  your data. Transparency is the ultimate currency
• 16:44
  right now, and Anthropix spent a massive amount
• 16:46
  of that currency this week. They really painted
• 16:48
  themselves into a corner. How so? Well, by declaring
• 16:51
  the Mythos class too dangerous for public release
• 16:53
  earlier in the development cycle, they set an
• 16:56
  impossible standard for themselves. When they
• 16:58
  finally did release Fable 5, they had to prove
• 17:01
  to regulators that they could control it. Right.
• 17:03
  But in doing so, they scrambled the product's
• 17:06
  utility and alienated the very professionals
• 17:08
  who were eager to use it. Some voices in the
• 17:11
  industry are actually trying to defend the strategy.
• 17:13
  Matt Sush suggests that perhaps it's better to
• 17:16
  cast a wide net and catch more false positives
• 17:18
  in these early days and then slowly relax the
• 17:21
  guardrails as the technology matures. But I mean,
• 17:23
  that is a very generous interpretation of a flawed
• 17:27
  rollout. The release a broken filter now fix
• 17:30
  the false positives later approach might be acceptable
• 17:34
  for a free consumer app. Yeah, maybe. But when
• 17:37
  you're charging premium enterprise subscription
• 17:39
  rates for frontier level tools, users expect
• 17:42
  the software to know the mathematical difference
• 17:44
  between a hacker writing malware and a high school
• 17:47
  student researching cell biology. They certainly
• 17:49
  don't expect their queries to be secretly rerouted
• 17:52
  to Opus 4 .8 without a price adjustment or notification.
• 17:55
  It is a masterclass in how not to deploy a highly
• 17:58
  anticipated product. It really is. And it leaves
• 18:00
  the entire AI community waiting to see if Anthropic
• 18:03
  can untangle this mess without compromising their
• 18:06
  core safety mission. Before we wrap up, a quick
• 18:08
  thanks to Old Glory, an iconic music and sports
• 18:11
  fan merch store, perfect for the techie in your
• 18:13
  life. Use promo code SNARFL for 15 % off at oldglory
• 18:18
  .com. So let's look at the big picture of everything
• 18:21
  we've covered today. Yeah, let's do a quick recap.
• 18:23
  We've seen a massive misstep from Anthropic on
• 18:26
  three entirely connected fronts. First, they
• 18:30
  tried to covertly control their competitors by
• 18:32
  degrading Fable 5 for AI researchers. hiding
• 18:35
  the fact that they were falling back to legacy
• 18:37
  models to protect their own market share. Second,
• 18:40
  they over -corrected their safety protocols,
• 18:43
  utilizing crude data vectors that block basic
• 18:45
  high school science questions and standard cybersecurity
• 18:48
  best practices. And third, to enforce this hyper
• 18:51
  -vigilant safety net, they implemented a 30 -day
• 18:54
  data retention policy that spooked Microsoft's
• 18:57
  legal team so badly they banned their own employees
• 19:00
  from using the tool. It all connects back to
• 19:02
  the core tension of the current technological
• 19:04
  era, which is this tug of war between safety
• 19:07
  and usability. Anthropic is terrified of the
• 19:09
  worst -case scenarios, the bioweapons and the
• 19:13
  mass cyber attacks. But in their desperate attempt
• 19:16
  to mitigate those extreme edge cases, they've
• 19:19
  engineered a product that is simultaneously too
• 19:22
  restricted to be useful for the experts and too
• 19:25
  invasive to be trusted by corporate compliance
• 19:27
  teams. It really leaves you wondering what the
• 19:29
  path forward actually looks like, because hardware
• 19:32
  and software are moving at light speed right
• 19:34
  now, but if the gargoyles are this clunky and
• 19:36
  invasive, the speed doesn't even matter. Which
• 19:39
  leaves us with a final thought for you to mull
• 19:41
  over as you go about your day. We are constantly
• 19:44
  sold the narrative that artificial intelligence
• 19:46
  is going to democratize knowledge, level the
• 19:49
  playing field, and unleash unprecedented human
• 19:52
  potential. Right. But look closely at the mechanics
• 19:54
  of what happened this week. If the only way to
• 19:56
  make the next generation of software quote -unquote
• 19:59
  safe is to severely limit what everyday people
• 20:02
  are allowed to ask it, and to meticulously log
• 20:05
  and retain everything they type for up to two
• 20:07
  years, are we actually building tools to empower
• 20:10
  people? Or are we just building a highly monitored
• 20:13
  walled garden reserved for a verified few? That
• 20:16
  is a question we are all going to have to answer
• 20:18
  very soon. We're here to help at Snarple, so
• 20:21
  reach out on our website if you have any more
• 20:23
  questions. That's a wrap on today's top stories.
• 20:30
  Remember, tech doesn't have to be complicated,
• 20:32
  and we're here to help connect the dots. If you've
• 20:35
  got questions or want to learn more, visit us
• 20:37
  at snarful .com. Thanks for listening, and we'll
• 20:41
  catch you next time.