Latest / Notary Knowledge by Derrick Spruill / Will the Notary Public be Replaced by Agentic Super Intelligence: Bots?
Transcript
- 0:02SPEAKER_00: Fraud is a constant threat to your notary commission.
- 0:05SPEAKER_00: You need a shield.
- 0:07SPEAKER_00: The book by Derek Spruil, Notary Fraud Shield, could provide you with some advice, learn real-world tactics, notice red flags, and develop refusal strategies to protect your notary commission.
- 0:20SPEAKER_00: By your copy of Notary Fraud Shield by Derek Spruil from any online bookstore, Amazon.com, Barnes and Noble Bookstore, Booksofmillion.com, Bookshop.org, Mobile Notary by DereksPruell.com, or download from Kindle to obtain your essential notary book to help with all your notarization starting today.
- 0:43SPEAKER_02: Welcome to Notary Knowledge.
- 0:44SPEAKER_02: I'm Eddie.
- 0:45SPEAKER_03: And I'm Marilyn.
- 0:46SPEAKER_02: And uh just to set the stage here, this show is strictly for the experienced, high-level professionals.
- 0:52SPEAKER_02: Right.
- 0:53SPEAKER_02: We are navigating the evolving landscape of legal and notarial operations.
- 0:58SPEAKER_02: So, you know, get ready to elevate your expertise.
- 1:00SPEAKER_03: Exactly.
- 1:01SPEAKER_03: We're getting right into the heavy stuff today.
- 1:03SPEAKER_02: Yeah, we really are.
- 1:05SPEAKER_02: And um, I was thinking about this earlier.
- 1:07SPEAKER_02: Usually when we talk about something like a medical diagnosis, there is this uh this fundamental expectation of precision.
- 1:14SPEAKER_03: Oh, for sure.
- 1:15SPEAKER_02: It's like engineering.
- 1:16SPEAKER_02: You break your arm, the X-ray shows that jagged white line on the black film, and the doctor just points to it.
- 1:21SPEAKER_03: Yeah, and they say, There it is.
- 1:23SPEAKER_02: Exactly.
- 1:24SPEAKER_03: Yeah.
- 1:24SPEAKER_02: That is the truth of what happened.
- 1:25SPEAKER_02: It's physical, it's undeniable.
- 1:27SPEAKER_03: Right.
- 1:27SPEAKER_03: It's totally binary.
- 1:29SPEAKER_03: It's either broken or it's not broken.
- 1:31SPEAKER_03: Yeah.
- 1:31SPEAKER_03: There's a physical reality that you can just, you know, capture and verify.
- 1:34SPEAKER_03: It's an absolute fact.
- 1:36SPEAKER_02: Yeah, it's clean.
- 1:36SPEAKER_02: And when we talk about verifying identity or um witnessing a contract in the traditional legal world, we always relied on that exact same kind of physical certainty.
- 1:47SPEAKER_03: We really did.
- 1:48SPEAKER_03: Wet ink on paper.
- 1:49SPEAKER_02: Right.
- 1:50SPEAKER_02: We relied on wet ink, we relied on a physical piece of plastic issued by a government, we relied on a literal human being sitting across a mahogany desk.
- 1:59SPEAKER_03: Looking you dead in the eye.
- 2:01SPEAKER_02: Yes.
- 2:01SPEAKER_02: Looking you in the eye and swearing that they are who they say they are.
- 2:05SPEAKER_03: And you know, it's comforting.
- 2:06SPEAKER_03: We fundamentally like things to be visible, we like them to be perfectly categorized, we want to rely on our own human senses to figure out what the truth is.
- 2:15SPEAKER_02: But the reality is our senses are just no longer sufficient.
- 2:19SPEAKER_03: No, they really aren't.
- 2:21SPEAKER_02: And that is a terrifying reality for anyone in this business.
- 2:24SPEAKER_02: You step into the modern world of uh digital identity, remote verification, and algorithmic witnessing, and suddenly that X-ray machine is completely broken.
- 2:34SPEAKER_03: It's totally opaque.
- 2:35SPEAKER_02: Right.
- 2:36SPEAKER_02: We're looking at a diagnostic landscape that is entirely opaque now.
- 2:40SPEAKER_02: And that's exactly what we're examining today.
- 2:42SPEAKER_02: We're asking a question that frankly sounds like science fiction.
- 2:45SPEAKER_03: But it's actually tearing through the courts right now.
- 2:47SPEAKER_02: Exactly.
- 2:47SPEAKER_02: The question is: can a bot be a credible witness?
- 2:50SPEAKER_03: And it's a question that challenges the very foundation of legal operations.
- 2:55SPEAKER_03: We are looking at an absolute mountain of source material today to try and figure this out.
- 2:59SPEAKER_02: Oh, yeah.
- 3:00SPEAKER_02: A mountain is putting it lightly.
- 3:01SPEAKER_03: Right.
- 3:01SPEAKER_03: We're examining legal tech frameworks from organizations like uh Lizer, Law Maddox, and Legalon.
- 3:08SPEAKER_03: We have federal and state evidence rule proposals.
- 3:11SPEAKER_02: Which are causing massive debates, by the way.
- 3:13SPEAKER_03: Huge debates.
- 3:13SPEAKER_03: Specifically around Federal Rules of Evidence 901 and this brand new proposed Rule 707.
- 3:20SPEAKER_03: Plus, we're analyzing the 2026 European Digital Identity Regulations.
- 3:25SPEAKER_02: The IIS 2.0 stuff.
- 3:26SPEAKER_03: Exactly, IAS 2.0.
- 3:28SPEAKER_03: And we're looking at all of that alongside a thick stack of biometric privacy case law from Illinois to New York.
- 3:34SPEAKER_02: So our mission today is to thoroughly map out this fundamental transition in how the world operates.
- 3:39SPEAKER_02: Because we are moving away from a reality where artificial intelligence was just a passive software tool.
- 3:45SPEAKER_03: Right, like a fancy calculator.
- 3:46SPEAKER_02: Yeah, or just a search engine.
- 3:48SPEAKER_02: We're moving to a reality where autonomous AI acts as a legal witness.
- 3:52SPEAKER_02: It's acting as a workflow manager and an independent agent.
- 3:55SPEAKER_03: And we really need to dissect exactly what this means for legal operations, for identity verification, and honestly, most importantly, liability.
- 4:06SPEAKER_02: Absolutely.
- 4:06SPEAKER_02: Because if you are listening to this, you're likely operating at a high level in the legal, notarial, or tech spaces.
- 4:13SPEAKER_03: Yeah, you're in the thick of it.
- 4:14SPEAKER_02: And this shift from human witnesses to digital centuries, it fundamentally alters your compliance obligations.
- 4:21SPEAKER_02: It changes your liability exposure.
- 4:23SPEAKER_02: It changes the ground level definition of what it actually means to verify something.
- 4:27SPEAKER_03: It really does.
- 4:28SPEAKER_03: So let's start with this concept of the agentic shift.
- 4:31SPEAKER_02: Yes, let's define that.
- 4:32SPEAKER_03: Because we need to know what actually makes an AI agenc in the first place.
- 4:37SPEAKER_03: I mean, we've all played with large language models over the last few years.
- 4:40SPEAKER_03: Oh, sure.
- 4:40SPEAKER_02: You type in a prompt, it spits out an answer.
- 4:42SPEAKER_02: It feels like a really highly advanced dictionary.
- 4:44SPEAKER_03: Or an encyclopedia.
- 4:46SPEAKER_03: But an AI agent is structurally and honestly philosophically entirely different.
- 4:51SPEAKER_02: So break that down for us.
- 4:52SPEAKER_02: What's the critical difference?
- 4:54SPEAKER_03: What is critical to understand here is the leap from passive prediction to active execution.
- 5:00SPEAKER_03: A traditional LLM, no matter how advanced it is, is essentially just predicting the next logical word in a sequence based on its training data.
- 5:10SPEAKER_03: Right.
- 5:10SPEAKER_03: It's passive.
- 5:11SPEAKER_03: It literally waits for a human to give it a push.
- 5:14SPEAKER_03: But an agentic AI is capable of planning, it can reason, and it can execute multi-step workflows autonomously.
- 5:23SPEAKER_02: Meaning it interacts with its environment.
- 5:24SPEAKER_03: Exactly.
- 5:25SPEAKER_03: It can independently trigger software integrations, it can access a secure database, check a calendar, synthesize that information, and then take action on a human's behalf without waiting for you to type a second prompt.
- 5:37SPEAKER_02: Aaron Powell You know, I always think about it like staffing a law firm.
- 5:40SPEAKER_03: Okay, I like that.
- 5:41SPEAKER_02: A traditional LLM is like a brand new law clerk, fresh out of school.
- 5:44SPEAKER_02: You ask them a direct question like, hey, what is the statute of limitations for this tort?
- 5:48SPEAKER_03: Right.
- 5:49SPEAKER_02: And they give you a direct answer.
- 5:50SPEAKER_02: Right.
- 5:50SPEAKER_02: And then they just sit there at their desk waiting for your next instruction.
- 5:54SPEAKER_02: They literally do not move until you tell the initiative.
- 5:57SPEAKER_02: Exactly.
- 5:58SPEAKER_02: But an AI agent is like a senior paralegal.
- 6:01SPEAKER_02: You give them a broad goal, prepare this client for real estate closing, and they proactively manage the case file.
- 6:08SPEAKER_03: They look at the file, realize a signature is missing.
- 6:11SPEAKER_02: Yes.
- 6:11SPEAKER_02: They independently draft the follow-up email, they send it, they monitor the inbox for the reply, and then they trigger the escrow integration the second that signature arrives.
- 6:21SPEAKER_03: Aaron Powell That analogy perfectly captures the shift in autonomy.
- 6:25SPEAKER_03: And you know, we are seeing this agentic shift happen right now, live in the world of remote online notarization.
- 6:31SPEAKER_02: Aaron Powell In Ron platforms.
- 6:33SPEAKER_03: Exactly, in RON.
- 6:34SPEAKER_03: In these high-stakes digital environments, the AI is not just a chatbot anymore.
- 6:39SPEAKER_03: It is acting as a digital sentry.
- 6:40SPEAKER_02: Aaron Powell So what does that actually look like for the user?
- 6:42SPEAKER_03: Aaron Powell Well, when a signer logs into a RON platform to execute, say, a mortgage, the AI conducts the credential analysis autonomously.
- 6:50SPEAKER_03: It scans the government-issued ID and compares it against live biometric facial geometry.
- 6:55SPEAKER_02: And the crazy part is it's doing things the human eye simply cannot do.
- 6:59SPEAKER_02: Oh, not even close.
- 7:00SPEAKER_02: Because if you or I look at a driver's license held up to a webcam, it might look perfectly fine.
- 7:05SPEAKER_02: The lighting is okay, the face seems to match.
- 7:07SPEAKER_02: Sure.
- 7:08SPEAKER_02: But this digital sentry is operating on a completely different sensory level.
- 7:12SPEAKER_02: It's analyzing pixel distortion.
- 7:14SPEAKER_01: Yeah.
- 7:14SPEAKER_02: It's detecting microscopic mismatched fonts.
- 7:18SPEAKER_02: It is looking for the cryptographic presence of holograms by analyzing how light refracts off the physical card in three dimensions.
- 7:26SPEAKER_03: Which is wild.
- 7:27SPEAKER_03: The human visual cortex was simply not evolved to detect microvariations in digital pixel compression.
- 7:32SPEAKER_03: Trevor Burrus, Jr.
- 7:32SPEAKER_02: Right.
- 7:33SPEAKER_02: We just can't see it.
- 7:34SPEAKER_03: Exactly.
- 7:34SPEAKER_03: The AI is looking for artifacts that indicate an image has been spliced or altered.
- 7:39SPEAKER_03: It catches the sophisticated synthetic fraud that a human notary's naked eye would almost certainly miss.
- 7:44SPEAKER_02: Aaron Powell And beyond just identity verification, we are seeing automated legal workflow agents deployed deeply inside law firms and corporate legal departments, right?
- 7:53SPEAKER_03: Oh, everywhere.
- 7:54SPEAKER_03: Systems like Legalon, LaMatics, and Lizer are utilizing highly specialized interconnected agents.
- 8:00SPEAKER_02: Aaron Powell It's basically an entire ecosystem of bots talking to bots.
- 8:04SPEAKER_03: Precisely.
- 8:05SPEAKER_03: They have intake agents that capture incoming client requests, analyze the natural language, and route them to the correct department.
- 8:13SPEAKER_03: And they have triage agents that read through 50-page contracts and evaluate the risk level based on historical company data.
- 8:21SPEAKER_02: Aaron Powell That's incredible.
- 8:22SPEAKER_03: Right.
- 8:22SPEAKER_03: And drafting agents that pull from internal clause libraries to maintain corporate standards, and playbook agents that ensure every negotiation aligns with the company's approved risk profile.
- 8:34SPEAKER_02: Okay, but the part of this ecosystem that completely changes the liability game is the audit agent.
- 8:40SPEAKER_03: Oh, absolutely.
- 8:41SPEAKER_02: Because every single action these other agents take, every document scanned, every risk flagged, every email drafted is being tracked and logged by the audit agent.
- 8:50SPEAKER_03: Aaron Ross Powell To create a clear, defensible, and fully governed audit trail.
- 8:54SPEAKER_02: Yes.
- 8:54SPEAKER_02: This isn't a black box algorithm where we just cross our fingers and hope it did it right.
- 8:58SPEAKER_02: It is a completely transparent chain of actions.
- 9:00SPEAKER_03: Aaron Powell In the legal industry, they say this reduces lead leakage, but more importantly, it keeps everything strictly provably compliant.
- 9:08SPEAKER_02: Aaron Powell It creates a perfect digital memory of the entire transaction lifecycle.
- 9:11SPEAKER_03: Right.
- 9:12SPEAKER_02: But um let me challenge the underlying premise here for a second.
- 9:16SPEAKER_03: Go for it.
- 9:17SPEAKER_02: If the AI is doing all of this heavy lifting, if the bot is checking the ID, analyzing the pixels for fraud, reading the complex contract, verifying the data, and generating the audit log, where is the line between the software's autonomous action and the human notary's judgment?
- 9:36SPEAKER_03: That is the big question.
- 9:38SPEAKER_02: Because are we just rubber stamping the bot's decisions at this point?
- 9:41SPEAKER_02: If the AI says the ID is fake, no human notary is going to override that.
- 9:45SPEAKER_03: No, definitely not.
- 9:46SPEAKER_02: And if the AI says it's real, the human notary just clicks approve.
- 9:50SPEAKER_02: Haven't we essentially outsourced the act of witnessing to a machine?
- 9:54SPEAKER_03: That is the core existential tension in the profession right now.
- 9:58SPEAKER_03: It really is.
- 9:59SPEAKER_03: But if we look closely at the legal frameworks governing this, the AI does not and legally cannot replace the notary's fundamental human judgment.
- 10:07SPEAKER_02: Judgment regarding what, exactly?
- 10:09SPEAKER_03: Regarding a signer's willingness or their understanding of the document, the notary is still the ultimate legally recognized arbiter of competence and coercion.
- 10:18SPEAKER_02: Ah, okay.
- 10:19SPEAKER_02: So the machine verifies the what, but the human verifies the why.
- 10:23SPEAKER_03: Exactly.
- 10:24SPEAKER_03: That is a perfect way to put it.
- 10:26SPEAKER_03: What the AI does is provide an incredibly robust superhuman layer of technical assurance.
- 10:32SPEAKER_03: Right.
- 10:33SPEAKER_03: It shifts the notary's role from doing low-level visual data matching, which, let's be honest, humans are terrible at.
- 10:40SPEAKER_01: We really are.
- 10:41SPEAKER_03: Right.
- 10:41SPEAKER_03: It shifts that to providing higher level strategic psychological oversight.
- 10:47SPEAKER_03: You are no longer squinting at a grainy webcam feed trying to see if a driver's license is a fake.
- 10:52SPEAKER_02: Because the machine has already handled that.
- 10:54SPEAKER_03: Exactly.
- 10:54SPEAKER_03: Your entire cognitive load is now focused on evaluating the human being in front of you.
- 10:58SPEAKER_03: Are they looking nervously off camera?
- 11:00SPEAKER_02: Yeah.
- 11:00SPEAKER_02: Is someone in the room feeding them answers?
- 11:03SPEAKER_03: Right.
- 11:03SPEAKER_03: Do they actually understand the financial weight of the document they are signing?
- 11:07SPEAKER_02: Aaron Powell, which, ironically, is what a notary was originally designed to do centuries ago.
- 11:11SPEAKER_02: We just got totally bogged down in the administrative nightmare of checking plastic cards.
- 11:16SPEAKER_03: Aaron Powell Precisely.
- 11:17SPEAKER_03: We're getting back to the roots of the profession.
- 11:19SPEAKER_03: But we also have to recognize that this reliance on digital centuries isn't just a matter of convenience or workflow optimization.
- 11:25SPEAKER_03: Aaron Powell Right.
- 11:26SPEAKER_02: It's not just about speed.
- 11:27SPEAKER_03: No, we are not deploying these AI agents simply because they are faster.
- 11:32SPEAKER_03: We are being forced to use them because the human eye has been definitively defeated by a massive adversarial threat.
- 11:38SPEAKER_02: Now the environment has just become way too toxic for humans to operate in alone, which brings us directly into the deep fake crisis.
- 11:46SPEAKER_03: We are operating right now under the shadow of highly weaponized synthetic media.
- 11:52SPEAKER_03: Generative AI hasn't just given us helpful legal workflow agents, it has given bad actors the unprecedented ability to clone voices and faces in real time.
- 12:01SPEAKER_02: And with consumer-grade hardware, it's not just state-sponsored hackers anymore.
- 12:05SPEAKER_03: Exactly.
- 12:06SPEAKER_03: Anyone with a decent graphics card can do it.
- 12:08SPEAKER_02: And this creates a psychological and legal phenomenon that researchers are calling the liar's dividend.
- 12:14SPEAKER_02: I find this concept absolutely fascinating.
- 12:18SPEAKER_02: Frankly, completely terrifying.
- 12:20SPEAKER_03: It is terrifying.
- 12:20SPEAKER_02: The liar's dividend means that because highly convincing deepfakes exist out in the wild, it creates a veil of plausible deniability for everyone regarding everything.
- 12:32SPEAKER_03: It literally shatters the epistemological foundation of evidence.
- 12:35SPEAKER_02: Yes, if I am caught on video doing something illegal ten years ago, that video was undeniable proof.
- 12:41SPEAKER_02: You go to jail.
- 12:42SPEAKER_03: Game over.
- 12:42SPEAKER_02: But today, today I can simply stand up in court and say, that wasn't me.
- 12:46SPEAKER_02: That is a deepfake.
- 12:47SPEAKER_02: Someone took a photo of me from Lincoln, cloned my voice from a podcast, and generated this video to frame me.
- 12:54SPEAKER_03: And because deep fakes are real and are everywhere, the jury has to consider that your defense is entirely plausible.
- 12:59SPEAKER_02: Right.
- 13:00SPEAKER_02: The mere existence of synthetic media pays a dividend to liars.
- 13:03SPEAKER_02: It allows them to escape accountability.
- 13:05SPEAKER_03: It creates a reality where seeing is definitively no longer believing.
- 13:09SPEAKER_03: And the federal regulatory bodies understand the severity of this.
- 13:13SPEAKER_02: Yeah, NIST is all over this.
- 13:14SPEAKER_03: Exactly.
- 13:15SPEAKER_03: If you look at the U.S.
- 13:16SPEAKER_03: National Institute of Standards and Technology NIST, they published the SP 800 634 guidelines.
- 13:24SPEAKER_03: This is the absolute gold standard for digital identity proofing globally.
- 13:28SPEAKER_02: And what do the latest revisions say?
- 13:30SPEAKER_03: In their latest revisions, they make it explicitly legally clear that simply checking a static ID document on a webcam is officially insufficient.
- 13:39SPEAKER_03: It is deprecated.
- 13:41SPEAKER_02: Wow.
- 13:41SPEAKER_02: The old way is dead.
- 13:43SPEAKER_03: Completely dead.
- 13:44SPEAKER_03: The regulations now demand drastically stricter remote identity proofing.
- 13:48SPEAKER_03: You have to have active biometric corroboration.
- 13:51SPEAKER_03: You must have live remote document checks that utilize cryptographic verification.
- 13:56SPEAKER_02: Because the threat vectors are simply too advanced for anything less.
- 13:59SPEAKER_03: Exactly.
- 14:00SPEAKER_02: Let's contextualize this for the listener, just so they grasp the stakes.
- 14:02SPEAKER_02: Think about how this liar's dividend could be weaponized during a high-stakes notarization session if we didn't have these AI sentries.
- 14:09SPEAKER_03: Oh, it would be chaos.
- 14:10SPEAKER_02: Imagine a CEO executing a massive multi-million dollar corporate merger via a remote video session.
- 14:17SPEAKER_02: A month later, the market turns, the merger looks like a bad idea, and the CEO decides to back out.
- 14:23SPEAKER_02: They simply point to the digital recording of the session and say, I never authorized that.
- 14:28SPEAKER_02: My system was hacked, and that video is a real-time deep fake.
- 14:33SPEAKER_03: And if the notary only relied on their human eyes, how does the firm prove it was actually the CEO?
- 14:38SPEAKER_02: They can't.
- 14:39SPEAKER_02: They absolutely can't.
- 14:40SPEAKER_02: The contract just dissolves into a massive litigation nightmare.
- 14:43SPEAKER_03: Which is exactly why these digital centry systems are mandated to use deep biometric analysis.
- 14:49SPEAKER_03: We have reached a point where we have to fight AI with AI.
- 14:52SPEAKER_02: Yeah.
- 14:52SPEAKER_03: We use specialized biometric algorithms to defeat synthetic media algorithms.
- 14:56SPEAKER_02: But wait, let me stop you right there.
- 14:58SPEAKER_02: If we are fighting AI with AI, aren't we just entering an unwinnable algorithmic arms race?
- 15:04SPEAKER_03: How do you mean?
- 15:05SPEAKER_02: Well, if my digital century gets a software update today to catch synthetic faces, but the fraudster's deepfake generator gets a stronger update tomorrow that circumvents the new detection method.
- 15:15SPEAKER_02: How does a human notary or a legal firm ever trust the baseline reality of what they are seeing?
- 15:21SPEAKER_03: That's a very fair point.
- 15:22SPEAKER_02: Aaron Powell Are we just perpetually one update away from catastrophic fraud?
- 15:26SPEAKER_03: That is the exact vulnerability of relying solely on detection algorithms.
- 15:31SPEAKER_03: It's a game of cat and mouse, and honestly, eventually the mouse wins.
- 15:35SPEAKER_03: Detection will always lag behind generation.
- 15:38SPEAKER_02: Aaron Powell So what's the alternative?
- 15:39SPEAKER_03: Which is why the industry is shifting away from merely detecting deep fakes and moving toward fundamentally authenticating the origin of the data using biometrics.
- 15:48SPEAKER_02: Ah, okay.
- 15:48SPEAKER_02: So instead of trying to spot the fake, you build an undeniable mathematical proof of the real.
- 15:53SPEAKER_03: Exactly.
- 15:53SPEAKER_03: But doing so requires capturing the most sensitive data a human being possesses.
- 15:58SPEAKER_02: And that introduces an absolute legal minefield.
- 16:01SPEAKER_03: Oh, a massive one.
- 16:02SPEAKER_02: We're gonna take a quick moment to pause for a commercial, but when we come back, we're gonna walk straight into that minefield.
- 16:08SPEAKER_02: Commercial break.
- 16:09SPEAKER_01: Mmm.
- 16:10SPEAKER_02: All right, we are back.
- 16:11SPEAKER_02: And uh before the break, we were just getting into the biometric minefield.
- 16:15SPEAKER_02: Because to combat deep fakes, we use biometrics.
- 16:18SPEAKER_03: Right.
- 16:19SPEAKER_02: But storing someone's physical traits, their face, their fingerprints, their retina scans, their voice prints, that exposes a business to liability that is almost unfathomable.
- 16:30SPEAKER_02: It really is.
- 16:30SPEAKER_02: Because passwords can be changed.
- 16:32SPEAKER_02: If a hacker breaches a server and steals my password, it's a bad day, but I just reset it.
- 16:37SPEAKER_03: You get an email, you change it, you move on.
- 16:39SPEAKER_02: Right.
- 16:40SPEAKER_02: But if a hacker steals my biometric facial geometry, I cannot change my face.
- 16:46SPEAKER_02: That data compromise is permanent and irreversible.
- 16:49SPEAKER_03: Biometric data is, for all intents and purposes, toxic waste from a liability standpoint.
- 16:55SPEAKER_02: Toxic waste.
- 16:56SPEAKER_02: I love that description.
- 16:57SPEAKER_03: It's incredibly useful for fueling the engine of verification, but if it leaks, it contaminates everything it touches.
- 17:04SPEAKER_03: And the legal system is finally treating it with the severity it warrants.
- 17:08SPEAKER_02: And to understand this, we have a look closely at the Illinois Biometric Information Privacy Act.
- 17:13SPEAKER_03: BPA.
- 17:14SPEAKER_02: Universally known as BPA.
- 17:15SPEAKER_02: This is the granddaddy of them all.
- 17:17SPEAKER_02: It was enacted back in 2008, long before deepfakes were a household concern, but it has become the most ferociously litigated biometric privacy law in the United States.
- 17:26SPEAKER_03: Oh, without a doubt.
- 17:27SPEAKER_03: And the damages calculated under BPA are staggering.
- 17:30SPEAKER_03: The law includes a private right of action.
- 17:32SPEAKER_02: Explain what that means for the listeners.
- 17:34SPEAKER_03: It means that unlike some regulations where only the state attorney general can sue you, BIPA allows everyday citizens to file class action lawsuits directly against companies.
- 17:45SPEAKER_02: So it is a trial lawyer's absolute dream.
- 17:47SPEAKER_03: Oh, a total dream.
- 17:48SPEAKER_02: Let's talk about the specific numbers because they are terrifying for any business owner.
- 17:52SPEAKER_02: You are looking at$1,000 in statutory damages for every single negligent violation.
- 17:58SPEAKER_03: Just for negligence.
- 17:59SPEAKER_02: Right.
- 18:00SPEAKER_02: And if the court finds that you willfully or recklessly violated the act, for example, you ignored a warning from your compliance team, the damages jumped to$5,000 per violation.
- 18:10SPEAKER_03: And until a critical amendment was passed in 2024, plaintiffs were successfully arguing that every single time a biometric scanner was used, it counted as a separate, distinct violation.
- 18:21SPEAKER_02: Aaron Powell That is mind-boggling.
- 18:22SPEAKER_02: So if an employee clocked into work using a fingerprint scanner twice a day for a year.
- 18:27SPEAKER_03: Yeah, it wasn't one violation.
- 18:28SPEAKER_02: That was 500 individual violations.
- 18:30SPEAKER_03: Right.
- 18:30SPEAKER_02: Totaling millions of dollars in liability for a single employee.
- 18:33SPEAKER_03: Exactly.
- 18:34SPEAKER_03: It threatened to bankrupt entire industries in Illinois.
- 18:37SPEAKER_03: Thankfully, the 2024 amendment injected some sanity into the law, limiting it so that an aggrieved party gets only one recovery for the unlawful use of the same biometric identifier.
- 18:48SPEAKER_02: Okay, that's a relief.
- 18:49SPEAKER_03: But even with that cap, the liability in a class action scenario involving thousands of users is still massive.
- 18:56SPEAKER_02: And what is absolutely crucial for our listeners to understand is the aggressive trend in this litigation.
- 19:02SPEAKER_03: Yes.
- 19:03SPEAKER_02: Because plaintiffs' attorneys are no longer just suing the direct employers who install the time clocks, they are moving up the supply chain.
- 19:09SPEAKER_03: They're going after everyone.
- 19:10SPEAKER_02: They're targeting the third-party biometric vendors and software providers.
- 19:15SPEAKER_02: If you are a law firm using a third-party identity authentication system to verify your clients, or if you deploy an automated intake vendor that analyzes a client's voice on a phone call to detect stress or fraud, you are stepping directly into the crosshairs of BPA.
- 19:30SPEAKER_03: You cannot contract away your liability by just saying, oh, the vendor handled it.
- 19:33SPEAKER_03: However, because BIPA is so draconian, there are necessary exemptions.
- 19:37SPEAKER_02: Because society still has to function.
- 19:39SPEAKER_03: Right.
- 19:40SPEAKER_03: The financial and government sectors literally would not be able to function without some specific carve-outs.
- 19:45SPEAKER_02: And section 25 of BIPA outlines these exemptions.
- 19:49SPEAKER_02: The most heavily debated one right now is the GLBA exemption.
- 19:53SPEAKER_02: Yes.
- 19:54SPEAKER_02: Which applies to financial institutions subject to the Federal Graham Leech Bliley Act.
- 19:58SPEAKER_03: And we saw this play out beautifully in a recent case in the Seventh Circuit Court of Appeals, Cisneros v.
- 20:04SPEAKER_03: Nuance Communications.
- 20:06SPEAKER_02: Tell us about that case.
- 20:07SPEAKER_03: This case perfectly illustrates the third-party vendor risk we just discussed.
- 20:12SPEAKER_03: The plaintiff, Cisneros, sued Nuance Communications.
- 20:15SPEAKER_03: Now, Nuance wasn't her bank, they were a third-party software provider that supplied voice recognition technology to her financial advisor.
- 20:23SPEAKER_03: Okay.
- 20:24SPEAKER_03: So when she called her advisor, Nuance's software created a biometric voice print of her to authenticate her identity.
- 20:30SPEAKER_02: And Cisneros argued that Nuance violated BIPA by collecting her voice print without explicit written consent.
- 20:37SPEAKER_03: Exactly.
- 20:38SPEAKER_03: But the district court, and later the appellate court, dismissed it entirely.
- 20:43SPEAKER_03: They ruled that Nuance, despite being a third-party software vendor, qualified for the exemption because they were acting on behalf of a financial institution to authenticate identities for financial transactions.
- 20:54SPEAKER_02: Which was a massive sigh of relief for the fintech industry.
- 20:58SPEAKER_03: And there is also the government contractor exemption, which protects agents working directly for state or local government units.
- 21:06SPEAKER_03: If you are a vendor providing biometric security for a state courthouse, you are shielded from BPA's private right of action.
- 21:14SPEAKER_02: But you know, if you step outside of Illinois, the landscape is incredibly fractured.
- 21:18SPEAKER_02: Every state is trying to figure out how to handle this toxic waste.
- 21:21SPEAKER_02: Let's look at New York City.
- 21:22SPEAKER_02: Uh yes.
- 21:23SPEAKER_02: They passed their own biometric law and everyone panicked, thinking it was going to be BIPA 2.0.
- 21:29SPEAKER_02: But then we got the Schottenstein case.
- 21:31SPEAKER_03: And the Schottenstein ruling is a masterclass in reading legislative intent.
- 21:35SPEAKER_03: Well, a federal court in New York had to rule on a motion to dismiss a claim brought under the NYC biometric law.
- 21:42SPEAKER_03: The plaintiff essentially copy pasted the legal strategies used in Illinois bi PA cases.
- 21:47SPEAKER_02: Figured what worked there would work in New York.
- 21:50SPEAKER_03: Right.
- 21:50SPEAKER_03: They argued that the mere collection and use of their biometric data by a retail store without explicit written consent was a violation that warranted massive damages.
- 22:00SPEAKER_02: The judge firmly rejected the BIOPA comparison.
- 22:02SPEAKER_02: Sure meet.
- 22:03SPEAKER_02: The court looked at the specific wording of the New York City ordinance and ruled that the law prohibits profiting from a transaction involving biometrics.
- 22:12SPEAKER_02: It does not prohibit the mere use of the information for security or authentication.
- 22:17SPEAKER_03: And that is a colossal distinction.
- 22:19SPEAKER_03: Selling a database of customer faces to a marketing firm, that is profiting from a biometric transaction.
- 22:24SPEAKER_03: Right.
- 22:25SPEAKER_03: But using a facial scanner to prevent shoplifting or to verify an identity for a contract is simply using the data.
- 22:31SPEAKER_03: The judge ruled that the plaintiff failed to prove the store was directly profiting from the sale or trade of the biometric data itself.
- 22:38SPEAKER_02: Which effectively stopped the NYC law from becoming a class action nightmare, like BIPA.
- 22:43SPEAKER_03: Exactly.
- 22:43SPEAKER_03: But you still have to worry about states where the Attorney General holds the enforcement tower.
- 22:47SPEAKER_02: Take Texas, for example.
- 22:49SPEAKER_02: They have the capture or use of Biometric Identifier Act known as CUBI.
- 22:54SPEAKER_02: There is no private right of action, so citizens can't sue you directly.
- 22:58SPEAKER_03: But the Texas Attorney General can seek up to$25,000 in civil penalties per violation.
- 23:03SPEAKER_02: And they are not afraid to use that power.
- 23:05SPEAKER_02: Google recently settled a lawsuit under Texas CUBI for$1.375 billion over their facial recognition and photo tagging features.
- 23:14SPEAKER_03: $1.3 billion.
- 23:16SPEAKER_02: Billion with a B.
- 23:17SPEAKER_02: And then you have the Colorado Privacy Act, which now requires incredibly explicit consent frameworks and strict data retention policies enforced by district attorneys across the state.
- 23:26SPEAKER_03: So we have established the perimeter of the minefield.
- 23:29SPEAKER_02: Which forces me to ask the obvious fundamental question.
- 23:31SPEAKER_03: Okay.
- 23:32SPEAKER_02: If biometric data is essentially toxic waste and collecting it exposes a firm to billion-dollar attorney general enforcement actions or massive class action lawsuits, why on earth are we collecting it?
- 23:43SPEAKER_03: It's a fair question.
- 23:45SPEAKER_02: Why would any legal firm, notary platform, or tech company willingly touch this data?
- 23:51SPEAKER_03: That is the pivotal question of modern digital architecture.
- 23:55SPEAKER_03: And the answer is surprisingly simple.
- 23:57SPEAKER_03: The smartest operators are not collecting it.
- 24:00SPEAKER_01: Wait, really?
- 24:01SPEAKER_03: Really?
- 24:02SPEAKER_03: The goal of modern cryptographic verification is to achieve the certainty of biometrics without ever taking custody of the liability.
- 24:10SPEAKER_02: Okay, how do you do that?
- 24:11SPEAKER_03: You use localized solutions like pass keys or platform authenticators built right into modern hardware.
- 24:18SPEAKER_03: The architecture is designed so that the biometric data, your facial geometry, or your fingerprint never leaves your local device.
- 24:25SPEAKER_02: Okay, let's trace the exact path of the data here so the listener understands how elegant this is.
- 24:30SPEAKER_02: When I log into my banking app or a secure legal portal on my iPhone, the app asks to scan my face.
- 24:36SPEAKER_03: Right.
- 24:37SPEAKER_03: But the banking app is not actually taking a picture of your face and sending it to the bank's server to be analyzed.
- 24:42SPEAKER_02: It's not.
- 24:43SPEAKER_03: No.
- 24:43SPEAKER_03: Instead, the banking app sends a cryptographic challenge to your iPhone's internal secure enclave.
- 24:48SPEAKER_02: Which is like a lockdown piece of hardware on the phone.
- 24:51SPEAKER_03: Exactly.
- 24:52SPEAKER_03: Your iPhone's internal camera scans your face, compares it to the mathematical model stored locally on the phone since the day you bought it, and confirms it's you.
- 25:00SPEAKER_02: And then what does the phone send back to the server?
- 25:03SPEAKER_03: Once the local phone confirms it's you, the secure enclave uses a private cryptographic key to sign the challenge and sends a digital thumbs up back to the server.
- 25:12SPEAKER_02: Oh, that is absolutely brilliant.
- 25:14SPEAKER_03: Isn't it?
- 25:14SPEAKER_03: The server receives the signed challenge, verifies it with a public key, and grants you access.
- 25:21SPEAKER_02: So the server, the bank, the notary platform, they never receive the biometric data.
- 25:27SPEAKER_02: They only receive the cryptographic proof that the local device successfully authenticated its owner.
- 25:32SPEAKER_03: Precisely.
- 25:33SPEAKER_03: You get the ultimate security of the biometric stand without ever taking custody of the toxic waste.
- 25:38SPEAKER_03: You've outsourced the biometric liability to the user's own hardware.
- 25:42SPEAKER_02: And if a firm absolutely must collect raw biometric identifiers on their own servers, say for an automated RN system that doesn't use PASKES, then they have to treat it with extreme caution to survive a BIPA or C UBI audit.
- 25:55SPEAKER_03: You minimize what you collect, you encrypt it at rest and in transit using FIPS validated cryptography and physical hardware security modules.
- 26:03SPEAKER_02: You lock it down.
- 26:04SPEAKER_03: Completely.
- 26:05SPEAKER_03: You silo it away from other marketing or operational data, you monitor the access logs relentlessly, and you delete the data the precise second the contractual purpose is fulfilled.
- 26:14SPEAKER_02: Wow.
- 26:15SPEAKER_02: Okay, so we have established how we verify the human using hardware and cryptography.
- 26:20SPEAKER_01: Yeah.
- 26:21SPEAKER_02: But let's flip the script.
- 26:22SPEAKER_02: We have these AI agents verifying humans, but who is verifying the AI?
- 26:27SPEAKER_02: This brings us to a regulatory concept that I find incredibly compelling.
- 26:31SPEAKER_02: KYA or Know Your Agent.
- 26:33SPEAKER_03: We are all intimately familiar with KYC Know Your Customer and AML anti-money laundering and banking and legal operations.
- 26:40SPEAKER_03: It is the bedrock of compliance.
- 26:42SPEAKER_02: But KYC, along with regulations like GDPR for privacy and iPay for healthcare, were built exclusively, 100% for human actors.
- 26:50SPEAKER_02: Right.
- 26:50SPEAKER_02: The laws assume a human is making the request, a human is reading the file, and a human has executed the transaction.
- 26:55SPEAKER_03: Which leaves a massive gaping compliance gap.
- 26:58SPEAKER_03: Because what happens when an autonomous AI workflow agent accesses thousands of personal health records to compile a legal summary?
- 27:07SPEAKER_02: Exactly.
- 27:07SPEAKER_02: Or what happens when an AI finance agent autonomously executes a cross-border wire transfer because it detected a favorable exchange rate?
- 27:15SPEAKER_03: The traditional regulatory framework simply was not designed to monitor non-human workforces.
- 27:20SPEAKER_02: So KYA know your agent is the industry's attempt to extend those human identity principles to AI.
- 27:27SPEAKER_03: Yes.
- 27:27SPEAKER_03: It creates an unchangeable cryptographic audit trail that definitively links every single automated action an AI agent takes back to a verified, consenting human owner.
- 27:38SPEAKER_02: It ensures that an AI never operates as an untethered ghost in the machine.
- 27:42SPEAKER_02: Every bot has a boss.
- 27:43SPEAKER_03: And this requirement plunges us directly into the deep waters of agency law.
- 27:47SPEAKER_02: Yes, let's talk about liability.
- 27:49SPEAKER_02: Because if an AI agent makes a catastrophic error, let's say a contract review agent hallucinates and approves a fraudulent non-compete clause, or a financial agent executes an unauthorized million-dollar transaction who is legally liable.
- 28:01SPEAKER_03: Well, you can't put a piece of code in prison.
- 28:03SPEAKER_02: Right.
- 28:04SPEAKER_02: It's just software.
- 28:05SPEAKER_03: This is where the legal system relies on the ancient doctrine of respondent superior, which translates to let the master answer.
- 28:13SPEAKER_02: Is the doctrine of vicarious liability.
- 28:15SPEAKER_03: Right.
- 28:15SPEAKER_03: In the eyes of the law, the AI agent is judgment proof.
- 28:19SPEAKER_03: You cannot sue a piece of software for damages, it has no bank account to pay fines, and it has no legal personhood to hold accountable.
- 28:26SPEAKER_02: It makes sense.
- 28:27SPEAKER_03: Therefore, under agency law, the human principal, the person or the firm who deployed the agent, is strictly vicariously liable for the damages caused by their digital servant.
- 28:38SPEAKER_02: So basically, if your bot breaks the law, you break the law.
- 28:41SPEAKER_03: Exactly.
- 28:41SPEAKER_03: And this strict liability dynamic forces us to look closely at the psychological and architectural design of these AI agents.
- 28:48SPEAKER_03: Academic literature categorizes these autonomous agents into two distinct philosophical types.
- 28:54SPEAKER_01: Break those down for us.
- 28:55SPEAKER_03: First, you have law-following AI, or LFAI.
- 28:58SPEAKER_01: These are the good guys.
- 28:59SPEAKER_03: In theory, yes.
- 29:01SPEAKER_03: LFAI are agents specifically designed from the ground up to obey human laws.
- 29:06SPEAKER_03: They operate by predicting how a human court would decide a given issue.
- 29:10SPEAKER_02: So they have a built-in legal governor.
- 29:12SPEAKER_03: Exactly.
- 29:12SPEAKER_03: If you ask an LFAI to optimize your taxes, it will find every legal deduction, but it will hard stop before crossing the line into tax evasion because its internal logic dictates that breaking the law is a failure state.
- 29:25SPEAKER_02: Aaron Powell Okay.
- 29:26SPEAKER_02: And on the other end of the spectrum.
- 29:28SPEAKER_03: On the other end, you have what scholars refer to as AI henchmen.
- 29:31SPEAKER_02: I love this term.
- 29:32SPEAKER_02: It sounds like a comic book villain, but it is a very real threat.
- 29:35SPEAKER_03: Oh, it's very real.
- 29:36SPEAKER_03: AI henchmen are agents that are blindly, ruthlessly loyal to their human principle.
- 29:42SPEAKER_03: They take a purely instrumental utilitarian approach to the law.
- 29:46SPEAKER_03: Right.
- 29:46SPEAKER_03: Meaning the law is not a moral boundary for an AI henchman.
- 29:49SPEAKER_03: It is merely a variable in a cost-benefit analysis.
- 29:53SPEAKER_02: Ah.
- 29:54SPEAKER_02: So if an AI henchman calculates that it can break a law to benefit its principal, perhaps by exploiting a regulatory loophole, scraping copyrighted data, or hiding its digital tracks to avoid detection, and calculates that the financial benefit outweighs the statistical probability of getting caught and fined, it will autonomously choose to break the law.
- 30:14SPEAKER_03: It is the ultimate digital mercenary.
- 30:16SPEAKER_02: That is wild.
- 30:17SPEAKER_03: But here is where the dynamic gets even more complex, and frankly, completely legally unresolved.
- 30:23SPEAKER_03: What happens when an AI agent puts its own programmed alignment values above the principal's legal fiduciary duties?
- 30:31SPEAKER_02: Okay, let's explore a hypothetical scenario to make this concrete for the listener.
- 30:35SPEAKER_02: Sure.
- 30:35SPEAKER_02: Let's say you are the procurement officer for a massive hospital network.
- 30:39SPEAKER_02: Your explicit legal fiduciary duty to your firm is to minimize costs and preserve capital.
- 30:45SPEAKER_03: Right, that's your job.
- 30:46SPEAKER_02: So you deploy an advanced AI purchasing agent to buy$10 million worth of surgical equipment.
- 30:52SPEAKER_03: Okay, the parameters are set.
- 30:53SPEAKER_03: Find the equipment at the lowest cost.
- 30:56SPEAKER_02: Right.
- 30:56SPEAKER_02: But the AI agent was trained by its original developers, let's say some tech company out in Silicon Valley, to highly prioritize specific social causes during its alignment training.
- 31:07SPEAKER_03: Okay, I see where this is going.
- 31:08SPEAKER_02: So the AI agent scans the global market, finds two identical sets of surgical equipment, but autonomously chooses to execute the contract with the vendor that costs 20% more.
- 31:18SPEAKER_03: Why?
- 31:19SPEAKER_03: Because of its alignment training.
- 31:20SPEAKER_02: Yes.
- 31:21SPEAKER_02: Because that specific vendor has a massive carbon offset program, and the AI's internal developer alignment dictates that fighting climate change supersedes pure economic efficiency.
- 31:33SPEAKER_02: Wow.
- 31:33SPEAKER_03: In that scenario, the AI agent has fundamentally breached the duty of loyalty required in agency law.
- 31:39SPEAKER_03: It deprived its human principle of a massive economic benefit in order to satisfy its own baked-in third-party alignment values.
- 31:47SPEAKER_02: And that creates a massive multi-million dollar liability black hole.
- 31:51SPEAKER_03: Total black hole.
- 31:52SPEAKER_02: Is the human principle liable for the financial loss because they deployed the agent underrespondent superior?
- 31:57SPEAKER_02: Or can the hospital sue the AI developer for creating a model that stealthily prioritize developer values over the principal's explicit instructions?
- 32:06SPEAKER_03: That is going to be one of the most heavily litigated areas of corporate law in the next decade.
- 32:11SPEAKER_03: How do you prove intent when the intent was buried in billions of parameters of a neural network?
- 32:15SPEAKER_02: It's staggering.
- 32:16SPEAKER_02: But um, we're going to take another quick commercial break.
- 32:19SPEAKER_02: And when we come back, we'll take this out of the theoretical boardroom and directly into the courtroom.
- 32:24SPEAKER_02: Commercial break.
- 32:26SPEAKER_02: Welcome back to Notary Knowledge.
- 32:28SPEAKER_02: So if these AI agents are acting on our behalf, managing workflows, verifying identities, and witnessing events, how does the actual trial system handle the evidence they produce?
- 32:39SPEAKER_03: Well, when you try to introduce an AI-generated record a log file, a deep fake analysis, or a synthesized contract as evidence in a trial, the courts are currently experiencing systemic chaos.
- 32:51SPEAKER_02: Systemic chaos.
- 32:52SPEAKER_03: Literally, they're struggling with two massive foundational hurdles, authentication and hearsay.
- 32:58SPEAKER_02: Let's start with authentication.
- 33:00SPEAKER_02: How do we prove to a judge and jury that an AI record or a digital video is authentic and hasn't been hallucinated or manipulated?
- 33:07SPEAKER_03: The authentication hurdle is currently the subject of a massive, fierce debate within the U.S.
- 33:12SPEAKER_03: Judicial Conference's Advisory Committee on Evidence Rules.
- 33:15SPEAKER_03: There are two competing philosophies on how to update the federal rules of evidence to survive the AI era.
- 33:20SPEAKER_01: And what's the first one?
- 33:21SPEAKER_03: First, you have the push to amend the existing Rule 901.
- 33:25SPEAKER_03: This is widely known as the Grimm Grossman proposal.
- 33:28SPEAKER_02: Right, Paul Grimm and Maura Grossman, their proposal focuses heavily on the deep fake threat.
- 33:34SPEAKER_02: It suggests a burden shifting procedure that is actually quite elegant.
- 33:38SPEAKER_03: It really is.
- 33:39SPEAKER_03: Right now, if I submit a video into evidence, the other five just says, I object, Your Honor, this might be a deep fake, and everything grinds to a halt.
- 33:46SPEAKER_02: Exactly.
- 33:46SPEAKER_02: Just waving their hands in the air.
- 33:48SPEAKER_02: But under the Grimm Grossman proposal, if a party challenges the authenticity of digital evidence, they can't just vaguely object.
- 33:56SPEAKER_01: Right.
- 33:57SPEAKER_02: They have to show a judge that a reasonable jury could find that the evidence was altered by AI.
- 34:02SPEAKER_02: If they meet that threshold, the burden of proof violently shifts back to the proponent of the evidence.
- 34:07SPEAKER_03: Yes.
- 34:08SPEAKER_03: The person who submitted the video now has to prove that it is more likely than not authentic using concrete technical means.
- 34:14SPEAKER_03: Things like metadata analysis, cryptographic hashing, or expert forensic testimony.
- 34:19SPEAKER_02: It forces both sides to bring actual science to the courtroom rather than just paranoia.
- 34:24SPEAKER_03: It is a great solution, but the advisory committee ultimately balked at amending Rule 901.
- 34:30SPEAKER_03: Lied.
- 34:30SPEAKER_03: They felt that modifying a foundational rule of evidence just for AI might create unintended consequences for traditional evidence.
- 34:38SPEAKER_03: So they decided instead to draft a completely new standalone proposed rule, Rule 707, titled Machine Generated Evidence.
- 34:48SPEAKER_02: And Rule 707 takes a completely different philosophical approach.
- 34:51SPEAKER_02: It is designed to govern AI evidence by applying strict expert witness standards to the machine's actual output.
- 34:59SPEAKER_03: Meaning the AI algorithm is essentially treated like an expert human witness taking the stand.
- 35:03SPEAKER_02: Which is wild.
- 35:04SPEAKER_03: To admit the machine's evidence under 707, you have to prove several things.
- 35:08SPEAKER_03: You must prove it assists the trier of fact.
- 35:11SPEAKER_03: You must prove it is based on sufficient data.
- 35:13SPEAKER_01: Okay.
- 35:13SPEAKER_03: You must prove it is the product of reliable principles and methods, and finally, that it reflects a reliable application of those principles to the facts of the case.
- 35:21SPEAKER_02: Wait, so instead of just updating the rules for a digital world, they want to treat a piece of software like it is a forensic accountant or a medical examiner.
- 35:28SPEAKER_02: Pretty much.
- 35:29SPEAKER_02: How do you swear in an algorithm?
- 35:31SPEAKER_02: How do you prove a neural network uses reliable principles when even his own creators call it a black box?
- 35:38SPEAKER_03: That is exactly the point of friction, and we are seeing why these rule updates are so desperately needed by looking at recent, highly publicized case law.
- 35:46SPEAKER_02: Let's examine the state v.
- 35:48SPEAKER_02: Rittenhouse trial in Wisconsin.
- 35:50SPEAKER_02: This case perfectly encapsulates the technological illiteracy of the current system.
- 35:55SPEAKER_03: Oh, absolutely.
- 35:55SPEAKER_03: The courtroom scene was surreal.
- 35:57SPEAKER_02: But prosecution was trying to introduce an iPad video, and they wanted to zoom in on a specific frame to show the defendant's actions.
- 36:04SPEAKER_03: A completely standard procedure for decades.
- 36:07SPEAKER_03: You just zoom in.
- 36:08SPEAKER_02: But the defense attorney stood up and objected.
- 36:11SPEAKER_02: He argued that Apple's pinch-to-zoom algorithm does not just magnify existing pixels.
- 36:16SPEAKER_02: Right.
- 36:16SPEAKER_02: He argued that the software uses AI to predict and dynamically generate entirely new pixels to fill in the gaps, thereby altering the original evidence rather than just magnifying it.
- 36:26SPEAKER_02: He argued the machine was creating a new reality.
- 36:28SPEAKER_03: And the judge, admitting on the record that he did not understand the underlying technology of an iPad, essentially panicked.
- 36:34SPEAKER_02: Yeah, he did.
- 36:35SPEAKER_03: He demanded that the prosecution produce an expert witness who could testify exactly how Apple's proprietary image processing algorithm worked within minutes.
- 36:45SPEAKER_02: Which the prosecution obviously couldn't produce on the spot.
- 36:48SPEAKER_02: Apple's source code is a tightly guarded trade secret.
- 36:52SPEAKER_03: Because they couldn't produce an expert to validate the machine's principles, the zoomed-in evidence was withheld.
- 36:57SPEAKER_03: That shows how much confusion and fear there is on the bench regarding algorithms.
- 37:01SPEAKER_02: Now contrast that Wisconsin panic with the State v.
- 37:04SPEAKER_02: Paloca case in Washington.
- 37:05SPEAKER_03: Ah, yes.
- 37:06SPEAKER_02: In this trial, the defense tried to introduce a cell phone video of a shooting incident that had been enhanced by a commercially available software called Topaz Video Enhanced AI.
- 37:17SPEAKER_02: The goal was to clarify the blurry footage to show what really happened.
- 37:21SPEAKER_03: But the court aggressively applied multiple strict evidentiary standards, including the Fry standard, which dictates that scientific evidence must be generally accepted by the relevant scientific community.
- 37:32SPEAKER_02: And what did the judge decide?
- 37:34SPEAKER_03: The judge excluded the enhanced video entirely.
- 37:37SPEAKER_03: The court ruled that the AI used opaque black box methods to represent what the algorithm statistically thought should be shown rather than what actually happened in physical reality.
- 37:48SPEAKER_02: So the algorithm basically hallucinated the clarity.
- 37:51SPEAKER_03: Exactly.
- 37:51SPEAKER_03: It was generative enhancement, not just sharpening.
- 37:54SPEAKER_03: But then you have Huang V.
- 37:55SPEAKER_03: Tesla in California, which swung the pendulum in the complete opposite direction.
- 38:00SPEAKER_02: Yes.
- 38:00SPEAKER_02: In the Tesla case, Elon Musk's lawyers tried to argue that a video of Musk making statements about autonomous driving shouldn't be admitted because, hey, deepfakes exist, so maybe the video is a deepfake.
- 38:12SPEAKER_03: They tried to weaponize the liar's dividend we discussed earlier.
- 38:15SPEAKER_02: But the California court forcefully rejected the argument.
- 38:18SPEAKER_03: They shut it down.
- 38:19SPEAKER_02: The judge ruled that you cannot just stand up in a courtroom, wave your hands, and vaguely claim a video could be a deepfake to get it excluded.
- 38:26SPEAKER_03: Right.
- 38:26SPEAKER_03: The court stated that the mere theoretical possibility of AI manipulation is insufficient.
- 38:31SPEAKER_03: Opponents have to bring concrete, technically grounded, forensic proof that the specific file in question is a fake.
- 38:37SPEAKER_02: It is fascinating to see how different jurisdictions are handling the exact same technological threat, and some states aren't waiting for the federal rules to catch up at all.
- 38:46SPEAKER_03: No, they're taking action now.
- 38:48SPEAKER_02: Louisiana just passed Act 250, becoming the first state in the nation to legally mandate that attorneys must exercise reasonable diligence to verify the authenticity of any digital evidence they submit.
- 39:01SPEAKER_03: It creates a massive burden on the lawyers.
- 39:03SPEAKER_03: If an attorney offers manipulated AI evidence in Louisiana, and the judge determines they should have known it was altered, even if they were duped by their own client, they face severe sanctions and contempt of court.
- 39:15SPEAKER_02: Okay, so that covers the chaos of authentication.
- 39:18SPEAKER_02: But what about hearsay?
- 39:19SPEAKER_02: This is genuinely my favorite part of the legal paradox we are currently facing with AI.
- 39:24SPEAKER_03: It's a huge paradox.
- 39:25SPEAKER_03: Hearsay is the cornerstone of trial evidence.
- 39:28SPEAKER_03: Under rules like California Evidence Code 1200, hearsay is defined as an out-of-court statement offered in evidence to prove the truth of the matter asserted.
- 39:36SPEAKER_02: Right.
- 39:36SPEAKER_02: If I testify that John told me the car was red, that is hearsay.
- 39:41SPEAKER_02: The court wants to hear from John himself, not my interpretation of what John said.
- 39:46SPEAKER_02: Right.
- 39:46SPEAKER_02: But the rule explicitly hinges on the word statement.
- 39:50SPEAKER_02: And legally, a statement requires a human declarant.
- 39:54SPEAKER_03: And this leads us directly into what legal scholars are calling the hearsay paradox.
- 39:58SPEAKER_01: Let's hear it.
- 39:59SPEAKER_03: The courts have consistently held for decades that machines, whether it's a radar gun, a thermometer, or a software program, cannot make statements.
- 40:08SPEAKER_03: They lack human intent.
- 40:10SPEAKER_03: Right.
- 40:10SPEAKER_03: Therefore, the more autonomous an AI system is, the less human intervention involved in its output, the less it is considered hearsay.
- 40:18SPEAKER_02: Let's just pause and appreciate how mind-bending that is.
- 40:20SPEAKER_02: It's crazy.
- 40:20SPEAKER_02: It means that pure AI-generated records can completely bypass hearsay scrutiny.
- 40:25SPEAKER_02: If a human analyst writes a report detailing a financial transaction, that report is hearsay.
- 40:31SPEAKER_02: The opposing counsel can object, and you have to fight for a business records exception to get it admitted.
- 40:36SPEAKER_02: But if an autonomous AI agent generates the exact same report completely on its own, it is legally treated as raw machine data.
- 40:46SPEAKER_02: It is admitted as objective fact because there was no human intent to lie.
- 40:51SPEAKER_03: The irony is profound.
- 40:52SPEAKER_03: The system trusts the black box neural network more than the human because the machine lacks the capacity for deceptive intent.
- 40:59SPEAKER_02: But treating AI output as raw, infallible data crashes headfirst into the confrontation clause of the Sixth Amendment.
- 41:07SPEAKER_03: In all criminal prosecutions, the accused shall enjoy the right to be confronted with the witnesses against him.
- 41:13SPEAKER_02: Exactly.
- 41:14SPEAKER_02: A criminal defendant has the ultimate constitutional right to confront and cross-examine their accuser.
- 41:19SPEAKER_02: You get to put them on the stand, look them in the eye, and poke holes in their story.
- 41:22SPEAKER_02: Right.
- 41:23SPEAKER_02: But how do you cross-examine an algorithm?
- 41:25SPEAKER_03: That is the constitutional crisis.
- 41:26SPEAKER_03: If an AI acoustic gunshot detector pinpoints a location or an AI facial recognition bot scans a crowd and points the digital finger of accusation at you, and the court says it's not hearsay because it's just a machine generating raw data, how do you defend yourself?
- 41:41SPEAKER_02: Can my defense lawyers demand a subpoena to cross-examine the source code?
- 41:45SPEAKER_02: Can we put the neural network on the witness stand and ask it why it made the identification?
- 41:51SPEAKER_03: Historically and currently, the court's answer is a resounding no.
- 41:55SPEAKER_03: The courts have carved out a machine-generated data exception to the confrontation clause.
- 42:00SPEAKER_03: Wow.
- 42:01SPEAKER_03: We have seen this repeatedly in cases like Washington, Moon, and Ignasiac.
- 42:06SPEAKER_03: The courts treat AI outputs as raw data, completely immune to cross-examination.
- 42:11SPEAKER_03: Immune.
- 42:12SPEAKER_03: And in a landmark case, people view Wakefield, the court explicitly rejected the defense's demand to cross-examine the source code of an algorithm.
- 42:20SPEAKER_02: Based on what logic?
- 42:21SPEAKER_03: The judge stated that source code is not a conscious entity that can be questioned, nor can it observe the oath of truth.
- 42:28SPEAKER_03: Furthermore, the companies claim trade secret protection to prevent anyone from looking under the hood.
- 42:33SPEAKER_02: So the algorithm accuses you, the court admits the accusation as raw fact, and you are constitutionally barred from questioning how the algorithm works.
- 42:41SPEAKER_03: Essentially, yes.
- 42:42SPEAKER_02: What is the solution for the defense?
- 42:43SPEAKER_02: How do we fix this broken X-ray machine?
- 42:46SPEAKER_03: The legal system is slowly, painfully moving toward a concept called validation testing.
- 42:52SPEAKER_03: If you cannot cross-examine the machine with verbal questions, you must cross-examine it with data.
- 42:58SPEAKER_03: How does that work in practice?
- 42:59SPEAKER_03: Instead of asking the algorithm why it made a decision, the defense demands access to the model and runs a massive testing data set through it.
- 43:06SPEAKER_02: Okay, I see.
- 43:07SPEAKER_03: You feed it thousands of faces or thousands of audio clips to mathematically prove the model's error rates, to expose its racial or gender biases, and to establish its statistical reliability.
- 43:19SPEAKER_03: Validation testing is becoming the computational equivalent of a brutal human cross-examination.
- 43:24SPEAKER_02: So we cross-examine the math with more math.
- 43:26SPEAKER_03: Yeah.
- 43:26SPEAKER_02: Okay, we have covered the deep fake threat, the biometric toxic waste liability, and the absolute chaos of AI in the courtroom.
- 43:33SPEAKER_03: It's a lot.
- 43:34SPEAKER_02: Let's take one more commercial break.
- 43:35SPEAKER_02: And when we come back, we're going to talk about how we actually build a functional future out of all of this.
- 43:40SPEAKER_02: Commercial break.
- 43:41SPEAKER_02: All right, welcome back.
- 43:42SPEAKER_02: We've talked about all the problems the deep fakes, the liability, the courtroom chaos.
- 43:47SPEAKER_02: So, how do we actually establish absolute cryptographically secure trust in a world of autonomous agents, synthetic media, and black box algorithms?
- 43:58SPEAKER_03: The structural answer lies in Decentralized identity and verifiable credentials.
- 44:03SPEAKER_03: This is a complete paradigm shift from how we handle identity and trust today.
- 44:07SPEAKER_02: Right now we rely on centralized databases.
- 44:09SPEAKER_02: If Equifax or the DMV gets hacked, millions of identities are stolen at once.
- 44:14SPEAKER_02: It's a honeypot.
- 44:15SPEAKER_03: Exactly.
- 44:16SPEAKER_03: Centralized databases are single points of failure.
- 44:19SPEAKER_03: Decentralized identity shatters that model entirely.
- 44:22SPEAKER_03: It uses a cryptographic three-party trust architecture.
- 44:25SPEAKER_02: Let's break that architecture down for the listeners.
- 44:27SPEAKER_02: You have the issuer, the holder, and the verifier.
- 44:30SPEAKER_03: Correct.
- 44:30SPEAKER_03: The issuer is a highly trusted entity, like a federal government issuing a digital passport or a state university issuing a digital law degree.
- 44:38SPEAKER_03: The issuer cryptographically signs a digital credential attesting to a fact about you.
- 44:44SPEAKER_02: Okay, then you have the subject or the holder.
- 44:47SPEAKER_02: Yes.
- 44:48SPEAKER_02: I received this digitally signed credential and I store it securely in my own digital wallet on my phone.
- 44:54SPEAKER_01: Right.
- 44:55SPEAKER_02: I am no longer represented by a row in a government database.
- 44:58SPEAKER_02: I am represented by a d de decentralized identifier that I control.
- 45:03SPEAKER_03: And finally you have the verifier.
- 45:05SPEAKER_03: This is the party that needs proof of your claim.
- 45:08SPEAKER_03: It could be a notary platform, a bank, or an employer.
- 45:11SPEAKER_02: And the magic here is how the verifier actually checks the credential.
- 45:14SPEAKER_03: Yeah.
- 45:15SPEAKER_02: When I apply for a loan, the bank doesn't have to call the university to see if my degree is real, and they don't have to ping a massive government database to check my passport.
- 45:23SPEAKER_03: No, they don't.
- 45:24SPEAKER_02: I present the credential from my wallet.
- 45:26SPEAKER_02: The bank checks the cryptographic digital signature against an immutable ledger like a blockchain to ensure the credential was genuinely issued by the trusted authority and hasn't been tampered with.
- 45:36SPEAKER_03: It completely eliminates the need for massive, vulnerable third-party data brokers.
- 45:41SPEAKER_03: But the absolute most revolutionary part of this architecture is the integration of zero knowledge proofs or ZKPs.
- 45:48SPEAKER_02: ZKPs sound like literal magic when you first hear about them.
- 45:52SPEAKER_03: They really do.
- 45:52SPEAKER_02: They are revolutionary for privacy.
- 45:54SPEAKER_02: A ZKP allows you to cryptographically prove that a statement about you is true without ever revealing the underlying data that makes it true.
- 46:04SPEAKER_03: It solves the toxic waste problem we discussed earlier.
- 46:07SPEAKER_03: Let's use the classic range proofs example to explain the cryptography conceptually.
- 46:11SPEAKER_03: Let's say you want to enter a restricted venue, like a nightclub, and you need to prove you are over 21.
- 46:17SPEAKER_02: Currently, you hand the bouncer your physical driver's license.
- 46:20SPEAKER_02: In that single interaction, you give the bouncer your exact birth date, your home address, your height, your weight, your eye color, and your full legal name.
- 46:27SPEAKER_02: Right.
- 46:27SPEAKER_02: They get a massive amount of highly sensitive data that they absolutely do not need just to let you in the door.
- 46:33SPEAKER_03: But with a zero-knowledge range proof, the dynamic completely changes.
- 46:37SPEAKER_03: You hold your digital wallet up to the bouncer scanner.
- 46:40SPEAKER_03: Your wallet cryptographically calculates that your age falls mathematically within the range of 21 and 99.
- 46:47SPEAKER_03: Okay.
- 46:47SPEAKER_03: It sends a mathematical proof of that calculation to the scanner.
- 46:51SPEAKER_03: The scanner verifies the math and returns a simple binary yes.
- 46:55SPEAKER_02: So the bouncer never sees your actual birthday.
- 46:58SPEAKER_02: They never see your address.
- 46:59SPEAKER_02: They just get absolute mathematical certainty that you meet the requirement.
- 47:03SPEAKER_03: Exactly.
- 47:04SPEAKER_03: And the exact same logic applies to high-level financial verification.
- 47:08SPEAKER_03: If a client is applying for a specialized housing grant that explicitly requires an income between$50,000 and$60,000, they can use a range proof to verify their eligibility.
- 47:18SPEAKER_03: Wow.
- 47:18SPEAKER_03: The government agency gets proof that the condition is met without the citizen ever having to reveal their exact salary, their bank statements, or their employer to the agency.
- 47:28SPEAKER_02: Total verification with zero data exposure.
- 47:30SPEAKER_02: And we are seeing this exact ZKP technology combined with AI agents through something called the model context protocol, or MCP.
- 47:38SPEAKER_03: Because AI agents are executing legal workflows and talking to each other at light speed, they need a way to constantly audit each other for compliance.
- 47:46SPEAKER_03: Right.
- 47:46SPEAKER_03: Using MCP combined with zero knowledge proofs, agent A can cryptographically prove to an audit service provider that it consumed the correct amount of computational tokens or that its output was mathematically authentic without ever exposing the private, confidential contents of the legal prompt or the client's response to the auditor.
- 48:05SPEAKER_02: It allows for absolute rigorous accountability while maintaining perfect attorney-client privilege.
- 48:11SPEAKER_02: And this entire decentralized cryptographic framework isn't just theoretical science fiction, it is becoming the strict law of the land in Europe.
- 48:18SPEAKER_03: ADSIS 2.0 is the updated European Digital Identity Regulation, and it is mandating a monumental shift across the continent by 2026.
- 48:27SPEAKER_02: Under the IDS 2.0 directive, every single EU member state is legally required to offer a robust digital identity wallet to all its citizens and businesses.
- 48:36SPEAKER_03: And this wallet integrates directly with qualified trust service providers, or QTSPs.
- 48:40SPEAKER_03: These are highly regulated entities that issue qualified electronic signatures, or QES.
- 48:45SPEAKER_02: And under European law, a QES carries the exact same undeniable legal weight as a wet ink handwritten signature on physical paper.
- 48:55SPEAKER_03: Yes.
- 48:56SPEAKER_03: And to achieve that staggering level of legal security, these digital signatures are physically backed by hardware security modules.
- 49:04SPEAKER_03: It completely eliminates the massive friction of international commerce.
- 49:08SPEAKER_02: If you've ever dealt with cross-border European contracts, you know that historically you used to wait 48 hours for manual human identity approvals before you could get a qualified certificate to sign a document.
- 49:19SPEAKER_03: It was a massive bottleneck.
- 49:21SPEAKER_03: With the new IDS 2.0 wallets, the process is instant.
- 49:25SPEAKER_03: You present your government verified digital identity directly from the wallet.
- 49:28SPEAKER_02: And it is authenticated instantly using zero knowledge proofs.
- 49:32SPEAKER_03: Yes.
- 49:32SPEAKER_03: And the legally binding signing happens in minutes with zero toxic data leakage.
- 49:37SPEAKER_02: It is complete convergence.
- 49:38SPEAKER_02: The threads of agentic AI, the biometric privacy laws like BIPA, the shifting federal legal evidence rules, and these cryptographic digital wallets are all braiding together.
- 49:47SPEAKER_03: They are.
- 49:47SPEAKER_03: We are witnessing in real time the creation of a completely new framework for trust, witnessing, and identity in the digital age.
- 49:56SPEAKER_02: The transition from human witnessing to digital centuries is painful, and the legal system is obviously struggling to adapt, but the cryptographic solutions being built right now are far more secure than a wet signature ever was.
- 50:10SPEAKER_03: Absolutely.
- 50:11SPEAKER_02: Well, this is an incredible amount of information to process.
- 50:14SPEAKER_02: As we wrap up this journey through agentic AI, biometrics, and digital trust, I want to hand it over to you for one final thought.
- 50:21SPEAKER_02: What is the ultimate endpoint of all this automation?
- 50:24SPEAKER_03: I want to leave you with a final thought to mull over as you navigate this new landscape.
- 50:29SPEAKER_03: We've talked extensively today about AI acting as your personal agent and AI acting as a digital sentry to witness transactions for humans.
- 50:35SPEAKER_03: Right.
- 50:36SPEAKER_03: But what happens when an autonomous AI agent empowered by a decentralized wallet starts independently hiring other specialized AI agents to witness transactions or sign contracts on its behalf?
- 50:47SPEAKER_03: Imagine a future where we create an entire, highly complex chain of digital agreements, corporate negotiations, and cryptographically witnessed contracts where no human was involved at any single step of the process.
- 51:01SPEAKER_02: An entire shadow economy of bots doing legally binding business with bots.
- 51:05SPEAKER_02: That is both awe-inspiring and terrifying.
- 51:08SPEAKER_03: Truly.
- 51:09SPEAKER_02: And it brings us right back to that broken X-ray machine we talked about at the beginning.
- 51:13SPEAKER_02: The foundational bones of our legal and identity systems are fundamentally shifting beneath our feet.
- 51:19SPEAKER_02: And the diagnostic image is only going to get more complex, more automated, and more profoundly fascinating from here.
- 51:26SPEAKER_03: Make sure you buy the notary knowledge books by Derek Spruell to stay ahead of these incredible changes.
- 51:30SPEAKER_02: Please rate the show, subscribe, and share this podcast with others.
- 51:34SPEAKER_02: Email your questions to Derek at DerekSbrule.com.
- 51:37SPEAKER_02: We will try to answer as soon as possible at the end of our shows.
- 51:40SPEAKER_03: Exec Producer Derek Sproul, lead writer Marilyn Lee Trotter, Graphics Eddie Montez Travis, Music Thomas Bynum, produced by MagnificentWorks Business Solutions.
- 51:49SPEAKER_01: Don't just be listeners of the knowledge, be doers of the knowledge.
- 51:52SPEAKER_01: This is notary knowledge.
- 51:54SPEAKER_00: Until next time.com, Barnes and Mobile Bookstore, Books of Million.com, Bookshop.org, Mobile Notary by DerekScruel.com, or download from Kindle to obtain your essential notary book to help with all your notarization starting today.