Latest / Connor with Honor: AI, Real Estate and Real Life / The Great American Artificial Intelligence Act of 2026

Transcript

The Great American Artificial Intelligence Act of 2026 · Connor with Honor: AI, Real Estate and Real Life

• 0:00
  SPEAKER_00: The Great American Artificial Intelligence Act of 2026, here it is.
• 0:04
  SPEAKER_00: We're going to give you the breakdown.
• 0:05
  SPEAKER_00: I'm going to go through it and we're going to talk about some of the highlights that I pulled out of it and think of it like Blade Runner, but without the robots.
• 0:13
  SPEAKER_00: At least not yet.
• 0:14
  SPEAKER_00: So we'll give you some ideas on what to expect.
• 0:17
  SPEAKER_00: Let's go through some of what we should be paying attention to and how this is going to apply as we go through it.
• 0:23
  SPEAKER_00: What I did is I worked with AI.
• 0:26
  SPEAKER_00: We've had some of the exact words highlighted yellow.
• 0:30
  SPEAKER_00: And of course, plainly we'll understand that.
• 0:33
  SPEAKER_00: The chip, top right, will have uh a neutral topic labor uh label.
• 0:38
  SPEAKER_00: So money, preemption, liability, these sorts of things.
• 0:41
  SPEAKER_00: Not a verdict, just kind of a kind of vanilla, if you will.
• 0:44
  SPEAKER_00: I'll get a little more nuanced inside of it.
• 0:46
  SPEAKER_00: And acronyms, of course, artificial intelligence.
• 0:49
  SPEAKER_00: IVO, something you might not have heard of at this point.
• 0:51
  SPEAKER_00: That's going to be independent verification organization, a private licensed AI auditor.
• 0:58
  SPEAKER_00: Interesting.
• 0:58
  SPEAKER_00: So that's uh going to be a new job title.
• 1:01
  SPEAKER_00: And they said AI was going to get rid of jobs.
• 1:03
  SPEAKER_00: Look at that.
• 1:03
  SPEAKER_00: We got more stuff.
• 1:04
  SPEAKER_00: The center, uh C-A-I-S-I.
• 1:07
  SPEAKER_00: Yeah, that's the acronym they came up with for this organization, Center for AI Standards and Innovation.
• 1:14
  SPEAKER_00: And that's a new office inside the commerce department.
• 1:18
  SPEAKER_00: How does one get a job there, I wonder?
• 1:20
  SPEAKER_00: I probably need some kind of college.
• 1:22
  SPEAKER_00: Yeah, what a bummer.
• 1:23
  SPEAKER_00: All right, so the director, head of the center, of course, Frontier Model, largest, most capable AI models, above a set compute threshold, of course, and they have all this laid out in here as well.
• 1:33
  SPEAKER_00: And we'll get a large frontier developer, uh, frontier model company with over$500 billion in revenue.
• 1:39
  SPEAKER_00: CBRN, Chemical, Biological, Radiological, or Nuclear, F-I-O-I-A, Freedom of Information Act, a public records law, and AG Attorney General, NIST, National Institutes of Standards and Technology.
• 1:53
  SPEAKER_00: Not to be confused with Standards and Time.
• 1:56
  SPEAKER_00: National Artificial Intelligence Research Research, N-A-I-R-R.
• 2:00
  SPEAKER_00: All right, let's move through.
• 2:01
  SPEAKER_00: So Title I, we're going to get into uh the first section of this.
• 2:04
  SPEAKER_00: If you haven't read this document, wherever I post this, I'll post up the document as well.
• 2:09
  SPEAKER_00: I don't know how how long it's going to live online.
• 2:12
  SPEAKER_00: I'll probably put it in a Google Doc and share that just so everybody has access as long as they want it or until they, you know, bury me in the Arizona Desert.
• 2:21
  SPEAKER_00: Frontier Artificial Intelligence Governance, where most of the real power sits.
• 2:25
  SPEAKER_00: A new regulator, new auditors, federal override to the states, and rewrites the fraud law, which will be exciting.
• 2:32
  SPEAKER_00: So catastrophic risk.
• 2:35
  SPEAKER_00: Easy to say, difficult to believe.
• 2:37
  SPEAKER_00: Title I, Section 1016 has a human number and a money number.
• 2:42
  SPEAKER_00: So this I thought was interesting.
• 2:43
  SPEAKER_00: 49 doesn't get there.
• 2:45
  SPEAKER_00: 50 people, though.
• 2:46
  SPEAKER_00: A foreseeable material risk and the death of serious injury to more than 50 people or more than a billion in damage to or loss of property as a result of a frontier model.
• 2:58
  SPEAKER_00: So 49 doesn't get there.
• 3:00
  SPEAKER_00: $999,999,999.99 is not there, but billion does.
• 3:07
  SPEAKER_00: The whole bill turns on this line.
• 3:09
  SPEAKER_00: This is right at the beginning when you read this thing.
• 3:11
  SPEAKER_00: Trigger for catastrophics 50 plus people killed or seriously hurt,$1 billion of property damage, human threshold and a dollar threshold put side by side.
• 3:27
  SPEAKER_00: Providing a chemical, biological, radiological nuclear weapon assistance that's not publicly available, engaging in conduct with no meaningful human oversight, that is a cyber attack, or would constitute murder, assault, extortion, or theft, evading control of such developer or a user.
• 3:47
  SPEAKER_00: Plainly, model only crosses the line if it helps to build chemical, biological, radiological, or nuclear CBRN weapon in ways you couldn't already find publicly, commits a serious crime on its own, no human steering, or slips the leash of its owner.
• 4:04
  SPEAKER_00: Probably not hard to see if that happens, if we're still left around to watch the fallout, no pun intended.
• 4:13
  SPEAKER_00: But trying to get there with our own human intellect, we're going to be very limited in trying to make these determinations.
• 4:20
  SPEAKER_00: So it's going to be very important that a very, very smart LLM personally assigned to each one of these agents, in some way gated from being corrupted or manipulated, is going to be part of this decision process.
• 4:35
  SPEAKER_00: It's part of the investigative layer.
• 4:38
  SPEAKER_00: All right, Title I section 1016B, but not if the government does it.
• 4:44
  SPEAKER_00: It doesn't include a foreseeable material risk from lawful activity of the federal government.
• 4:49
  SPEAKER_00: And did you expect anything else?
• 4:52
  SPEAKER_00: All right.
• 4:52
  SPEAKER_00: Plainly dangerous capability, different rules.
• 4:55
  SPEAKER_00: If the harm flows from a lawful federal government activity, it's carved out of the definition of catastrophic risk.
• 5:03
  SPEAKER_00: Yeah, they have to get their uh their excuses in there.
• 5:06
  SPEAKER_00: Title I, Section 102, the law leans on interpretability, understanding how AI actually thinks.
• 5:12
  SPEAKER_00: AI interpretability means the degree to which humans are able to accurately understand, explain, and evaluate how an artificial intelligence model processes information.
• 5:23
  SPEAKER_00: I don't think we know any of the answers to any of this stuff.
• 5:28
  SPEAKER_00: It's interesting it's here.
• 5:29
  SPEAKER_00: I guess definitions are so we all understand the playing field, understand the rules, including the extent to which humans can identify the cases of unexpected, incorrect, harmful, deceptive, or unintended model behavior.
• 5:43
  SPEAKER_00: The simple things, absolutely, doing something like it talked about and us not being okay with it, easy to see.
• 5:50
  SPEAKER_00: The more clever approaches to things, the multidimensional, the 3D chess stuff, yeah, it's gonna be interesting.
• 5:56
  SPEAKER_00: Plainly, the safety regimen assumes humans can understand and explain artificial intelligence models, reach their outputs.
• 6:04
  SPEAKER_00: Now, something the field of itself openly admits cannot fully do, arguably, probably less over time.
• 6:10
  SPEAKER_00: I'm gonna see less and less the ability to understand it.
• 6:12
  SPEAKER_00: And, and, here we go.
• 6:14
  SPEAKER_00: Let's let's get another one here.
• 6:16
  SPEAKER_00: And that we keep it from being convinced.
• 6:19
  SPEAKER_00: Adversarial robustness means the degree to which an artificial intelligence model is able to resist or withstand the exploitation of an AI system vulnerability while maintaining integrity, reliability, security, and privacy.
• 6:34
  SPEAKER_00: And the way that I read it, the bill's other big assumption is that artificial intelligence models can be reliably kept from being hacked, jailbroken, or talked into doing the wrong thing.
• 6:45
  SPEAKER_00: The cat-on-x that breaks, jailbreaks everything.
• 6:49
  SPEAKER_00: Yeah, that's that's an issue.
• 6:51
  SPEAKER_00: And big brain on that guy, probably bigger brains exist that might not have seen the excitement in doing those things.
• 6:59
  SPEAKER_00: Hopefully they never wake up.
• 7:01
  SPEAKER_00: That the gates and controls hold even as the models outpace the people writing the rules.
• 7:06
  SPEAKER_00: Again, they they're admitting it's predicated on our human intellect, and that's only gonna get us to about a cup of coffee and maybe half a Danish, and then it's gonna blow out from there.
• 7:17
  SPEAKER_00: All right.
• 7:18
  SPEAKER_00: Uh, Title I, Section 10113, who's actually covered a compute line in the sand?
• 7:23
  SPEAKER_00: Frontier model means a foundation model trained utilizing a quantity of computing power greater than the this is that uh teraflops thing, 10 10 to the 26th integer floating point operations.
• 7:37
  SPEAKER_00: I it's in the document.
• 7:39
  SPEAKER_00: There you go.
• 7:39
  SPEAKER_00: There's that.
• 7:40
  SPEAKER_00: I never thought I would see this in in uh in a government document.
• 7:44
  SPEAKER_00: Uh and again, no concept.
• 7:47
  SPEAKER_00: Somebody explain it to me, please.
• 7:49
  SPEAKER_00: But plainly, greater than 10 to the 26th power of operations, only the biggest models are frontier.
• 7:56
  SPEAKER_00: Draw the line here, and you decide who lives under this law and who walks free is a model with less.
• 8:05
  SPEAKER_00: 25th, 24th, 23rd.
• 8:09
  SPEAKER_00: The to that exponent regarding operations, is that dangerous or it doesn't fall underneath this bill?
• 8:15
  SPEAKER_00: Maybe.
• 8:17
  SPEAKER_00: Two money gates, these are going to decide how all the rules are going to hit you.
• 8:21
  SPEAKER_00: Frontier developer, more than 50 million in gross revenues.
• 8:25
  SPEAKER_00: Folks, these these numbers are remarkable.
• 8:27
  SPEAKER_00: Large frontier developer, more than 500 million in gross revenue.
• 8:32
  SPEAKER_00: Plainly, so over 50 million in revenue makes you a frontier developer, over 500 million.
• 8:37
  SPEAKER_00: These are more definitions.
• 8:38
  SPEAKER_00: Over 500 million makes you a large one.
• 8:41
  SPEAKER_00: And the heaviest duties, audits, frameworks fall only on the large club.
• 8:47
  SPEAKER_00: A brand new federal AI office is created, Title I, Section 102.
• 8:51
  SPEAKER_00: Secretary of Commerce, another another governmental organization shall establish a Center on Artificial Intelligence to be known as the Center for AI Standards and Innovation within 60 days after enactment.
• 9:02
  SPEAKER_00: Okay.
• 9:03
  SPEAKER_00: Short sleeves, ties, brown ties, maybe vertical stripes, short sleeve collared shirts, brown ties, maybe no stripes, maybe maybe white with a jacket, short sleeves and thick black rimmed glasses for everybody.
• 9:22
  SPEAKER_00: Pocket protectors have to be.
• 9:24
  SPEAKER_00: We're talking 60s tech dress.
• 9:28
  SPEAKER_00: I think that should be the standard uniform.
• 9:30
  SPEAKER_00: Very much men in black-ish.
• 9:32
  SPEAKER_00: I think that would be really good.
• 9:34
  SPEAKER_00: Men in black, yeah, men in black's good.
• 9:37
  SPEAKER_00: Uh the matrix, the agents on there.
• 9:40
  SPEAKER_00: Maybe not with as much pizzazz as the agents.
• 9:43
  SPEAKER_00: They had some pretty good tailoring going on.
• 9:44
  SPEAKER_00: So plainly within 60 days, the Commerce Department stands up a new office.
• 9:48
  SPEAKER_00: Center for AI Artificial Standards and Innovation.
• 9:51
  SPEAKER_00: How does a brother get on lot B for that one?
• 9:53
  SPEAKER_00: That runs the audits, reporting, and the auditor licensing.
• 9:57
  SPEAKER_00: Good stuff.
• 9:58
  SPEAKER_00: $300 million plus fees billed to the companies.
• 10:01
  SPEAKER_00: Another one.
• 10:01
  SPEAKER_00: Title I, Section 102.
• 10:05
  SPEAKER_00: $100 million for fiscal year 2027,$100 million for fiscal year 2028, and$100 million for 2029.
• 10:13
  SPEAKER_00: Director may establish and collect reasonable fees from large frontier developers.
• 10:18
  SPEAKER_00: Yeah, Trump is, well, 2029, I think that's it, right?
• 10:22
  SPEAKER_00: 2028.
• 10:22
  SPEAKER_00: So a little bit of overlap, but you know, but who knows what's going to happen at that particular point.
• 10:29
  SPEAKER_00: But I don't know how much time is left before these systems go excitingly more advanced.
• 10:36
  SPEAKER_00: Excitingly more advanced.
• 10:38
  SPEAKER_00: Plainly, taxpayer funds$100 million a year for three years.
• 10:41
  SPEAKER_00: $300 million total on top of that center can bill the auditors and the big AI companies' fees to cover their costs.
• 10:48
  SPEAKER_00: Of course they can.
• 10:50
  SPEAKER_00: Appearance of evil in that, them being able to build the auditors, what the auditors are going to charge to do their work.
• 10:57
  SPEAKER_00: Is there an appearance of evil there?
• 10:59
  SPEAKER_00: Yeah, interesting.
• 11:00
  SPEAKER_00: Maybe, maybe not.
• 11:01
  SPEAKER_00: We'll see.
• 11:02
  SPEAKER_00: It can pay its experts almost anything.
• 11:04
  SPEAKER_00: Secretary shall have the authority to fix the basic pay of such personnel at a rate, not to the X, not in excess of the highest total annual compensation payable under Section 104, Title III.
• 11:16
  SPEAKER_00: Normal federal pay caps are waived.
• 11:19
  SPEAKER_00: So you're not stuck at the poor FBI payment, which is atrociously low.
• 11:24
  SPEAKER_00: Or CIA or these others, you know, you think you think they just make a killing in these governmental organizations.
• 11:30
  SPEAKER_00: And this one, yeah, maybe, maybe there's a killing to come.
• 11:33
  SPEAKER_00: But in the other governmental agencies, even Border Patrol high GS rating, uh, pretty good starting pay.
• 11:38
  SPEAKER_00: But for the danger, is it really good?
• 11:40
  SPEAKER_00: Yeah, that's the thing.
• 11:41
  SPEAKER_00: You see the numbers, but then you're gonna be running at the bullet.
• 11:44
  SPEAKER_00: I don't know if that's a great thing.
• 11:45
  SPEAKER_00: I was LAPD for a long time, yeah, running at the bullet.
• 11:48
  SPEAKER_00: That's uh that's that's the way it should be.
• 11:51
  SPEAKER_00: But you see some cops on, I'm not gonna go there, folks.
• 11:54
  SPEAKER_00: I'm stopping.
• 11:54
  SPEAKER_00: The center can pay technical hires up to roughly vice president level compensation, and a prior 15-person staffing cap is deleted.
• 12:05
  SPEAKER_00: Interesting.
• 12:06
  SPEAKER_00: What companies uh share can't be used to regulate them.
• 12:10
  SPEAKER_00: Information shared with the center uh may not be made public or used by any federal, straight, local, or tribal government.
• 12:17
  SPEAKER_00: Got the Indians in there too to regulate an activity of such an entity.
• 12:22
  SPEAKER_00: Plainly, when a data a company voluntarily hands the center is shielded in two ways.
• 12:27
  SPEAKER_00: It stays secret from the public, and no government at any level may use it to regulate that same company.
• 12:32
  SPEAKER_00: We got some good whistleblower stuff going on in this system as well.
• 12:36
  SPEAKER_00: Nice to be had nice to have that have been thought of.
• 12:40
  SPEAKER_00: Big developers must publish a safety framework.
• 12:43
  SPEAKER_00: Large frontier developers shall write, implement, comply with, clearly and conspicuously publish on a publicly available website, Frontier Framework.
• 12:52
  SPEAKER_00: Within a year, every large frontier model developer must post a public document explaining how it tests for and manages catastrophic risk in the models.
• 13:01
  SPEAKER_00: Good.
• 13:02
  SPEAKER_00: All right.
• 13:02
  SPEAKER_00: No false statements unless they were reasonable.
• 13:04
  SPEAKER_00: This is cute.
• 13:05
  SPEAKER_00: A frontier developer may not knowingly make an inaccurate statement about or convey a false impression, exception.
• 13:12
  SPEAKER_00: Does not apply to a statement made in good faith that was reasonable under the circumstances.
• 13:16
  SPEAKER_00: That's good.
• 13:16
  SPEAKER_00: That's good to know.
• 13:17
  SPEAKER_00: Plainly, lying in a safety report is prohibited.
• 13:21
  SPEAKER_00: So you can't outright lie, but there's a built-in escape hatch for statements made in good faith that were reasonable, which is the kind of language lawyers like to fight over.
• 13:32
  SPEAKER_00: There's that one for me to you.
• 13:33
  SPEAKER_00: Companies redact their own safety disclosures.
• 13:36
  SPEAKER_00: A frontier development may make a redaction to protect a trade secret or the cybersecurity of such developer or public safety or the national security.
• 13:45
  SPEAKER_00: And what that means to me before anything is published, the company itself decides what to block out for trade secrets, cybersecurity, public safety, or national security, and only has to keep the hidden material for five years.
• 13:59
  SPEAKER_00: Yeah, there's that.
• 14:01
  SPEAKER_00: All right.
• 14:01
  SPEAKER_00: Imminent danger, 24 hours to call law enforcement.
• 14:05
  SPEAKER_00: Not later than 24 hours after a frontier developer discovers a critical safety incident that poses an immediate risk of death or serious physical injury.
• 14:15
  SPEAKER_00: Such developer shall report such an incident to law enforcement agency.
• 14:19
  SPEAKER_00: So 49 people, no, 50 people, yes, 999 million and change, no, 1 billion, yes.
• 14:27
  SPEAKER_00: Could it be?
• 14:28
  SPEAKER_00: Or any, any safety problem.
• 14:33
  SPEAKER_00: Imminent threat to life.
• 14:34
  SPEAKER_00: Company has 24 hours to notify law enforcement for other serious incidents.
• 14:39
  SPEAKER_00: It has 15 days to notify the center.
• 14:42
  SPEAKER_00: You know, maybe maybe that other one should be the same amount of time.
• 14:46
  SPEAKER_00: Maybe 1 million a day, a civil penalty of not more than 1 million per violation.
• 14:52
  SPEAKER_00: Each day during which a violation continues constitutes a separate violation.
• 14:57
  SPEAKER_00: So penalties stop out at a million bucks per violation.
• 15:00
  SPEAKER_00: Every single day it continues, counts as a new violation.
• 15:04
  SPEAKER_00: So those numbers, yeah, they get up pretty quick, pretty fast.
• 15:07
  SPEAKER_00: And when you're talking about trillions of dollars, yeah, you probably take a few weeks to get somewhere crazy.
• 15:12
  SPEAKER_00: State attorneys general can opt in to enforce.
• 15:15
  SPEAKER_00: State attorneys general may opt in to receive reports submitted to the director and bring action to collect a civil penalty.
• 15:21
  SPEAKER_00: So beyond the U.S.
• 15:22
  SPEAKER_00: Attorney General, the state attorney general who opts in to get confidential incident reports can sue companies to enforce the rules.
• 15:30
  SPEAKER_00: More teeth, new licensed job, an AI auditor, the IVO.
• 15:36
  SPEAKER_00: This is good.
• 15:36
  SPEAKER_00: Um I'm warming up for this one.
• 15:38
  SPEAKER_00: Large frontier developers shall retain independent verification organization, IVO, that's been licensed with independence from the artificial intelligence industry.
• 15:47
  SPEAKER_00: So we're talking standalone, no influence, no promises, no dinners, long walks on the beach, romantic interludes.
• 15:55
  SPEAKER_00: Yeah, super clean.
• 15:56
  SPEAKER_00: The bill invents a profession.
• 15:58
  SPEAKER_00: Independent verification organization, IVO.
• 16:02
  SPEAKER_00: I see it.
• 16:02
  SPEAKER_00: I see it.
• 16:03
  SPEAKER_00: Like I C E IVO.
• 16:04
  SPEAKER_00: I see it.
• 16:05
  SPEAKER_00: I'm seeing it.
• 16:06
  SPEAKER_00: Turn it upside down, and it really looks cryptic.
• 16:08
  SPEAKER_00: I like it.
• 16:09
  SPEAKER_00: Is a private government licensed firm that audits the AI companies.
• 16:14
  SPEAKER_00: Law requires it to be independent from the AI, artificial intelligence industry itself.
• 16:19
  SPEAKER_00: I'm I got the men in black thing going on here in my brain.
• 16:24
  SPEAKER_00: I want to apply.
• 16:25
  SPEAKER_00: I want to get my finger, my, my um, what are these things called?
• 16:30
  SPEAKER_00: Finger, not finger tips, fingerprints.
• 16:32
  SPEAKER_00: I want to get those burned off on that ball they have, and then get to see all the other exciting stuff inside.
• 16:38
  SPEAKER_00: Uh agent B, Agent K, Agent J, and all this other stuff.
• 16:42
  SPEAKER_00: Very excited about this.
• 16:43
  SPEAKER_00: We'll see.
• 16:44
  SPEAKER_00: Auditor gets to see everything, unredacted.
• 16:47
  SPEAKER_00: That's a lot for a human brain.
• 16:48
  SPEAKER_00: I was a cop for a long time, saw some horrible things.
• 16:51
  SPEAKER_00: Don't think it affected me, at least not that I would admit to.
• 16:55
  SPEAKER_00: But um, yeah, this is going to be that type of thing where they're gonna see some pretty powerful stuff, and if they're not psychologically prepared, or if they have leverage points that bad actors can get leverage and hooks into them.
• 17:13
  SPEAKER_00: Yeah.
• 17:13
  SPEAKER_00: If you don't have anybody, don't care about anybody, probably good position.
• 17:17
  SPEAKER_00: But if you have family and people you care about, yeah, there's some leverage there.
• 17:21
  SPEAKER_00: Talk to prison guards.
• 17:22
  SPEAKER_00: That's a rough job.
• 17:23
  SPEAKER_00: You're around Supermax people all day long, you're in the yard with them.
• 17:27
  SPEAKER_00: Yeah, it's a tough job being a prison guard.
• 17:29
  SPEAKER_00: Imagine at this scale, this isn't just in a prison where you have the uh the tribal uh law exist.
• 17:38
  SPEAKER_00: This is this is much, much bigger.
• 17:40
  SPEAKER_00: So watch out.
• 17:41
  SPEAKER_00: That hiring process needs to be needs to be pretty unique.
• 17:44
  SPEAKER_00: A large frontier developer shall grant an IVO timely access upon request to unredacted materials, records, personnel systems, and all other information reasonably necessary.
• 17:55
  SPEAKER_00: Yeah, I think this should be an armed position and not just the regular federal certification for armament, but I think I think these people need to carry guns.
• 18:04
  SPEAKER_00: I really, really do.
• 18:05
  SPEAKER_00: I think that's important.
• 18:07
  SPEAKER_00: Probably wear body armor under that nice late 60s tech suit.
• 18:11
  SPEAKER_00: Plainly, the independent verification organization, IVO, can demand the redacted files, the systems, and the people.
• 18:18
  SPEAKER_00: Deep level access to the most valuable secrets in the company.
• 18:22
  SPEAKER_00: That's something else.
• 18:23
  SPEAKER_00: You gotta have people that that that aren't gonna give up a secret, people that are gonna really believe the oath they took.
• 18:30
  SPEAKER_00: No bullshit people, people that are gonna keep their mouth shut, people that when pressed aren't going to give in.
• 18:36
  SPEAKER_00: I mean, this this goes on the same level as a cop running towards the person shooting the gun, not hiding behind the pole, but going and getting them.
• 18:47
  SPEAKER_00: This is gonna be a big deal.
• 18:49
  SPEAKER_00: This is torture-esque stuff.
• 18:51
  SPEAKER_00: This is dare I say CIA on steroids.
• 18:56
  SPEAKER_00: Present company excluded, of course.
• 18:58
  SPEAKER_00: I just work out a lot.
• 18:59
  SPEAKER_00: No, there's a little bit of TRT, but we're not gonna get into that.
• 19:02
  SPEAKER_00: Yeah, very interesting.
• 19:03
  SPEAKER_00: I like to see it.
• 19:03
  SPEAKER_00: The auditors are shielded from lawsuits.
• 19:06
  SPEAKER_00: IVO shall be immune from suit and liability under federal and state law.
• 19:10
  SPEAKER_00: Uh again, another big power play.
• 19:13
  SPEAKER_00: These people need to be selected.
• 19:14
  SPEAKER_00: I don't think a 20-year-old is gonna have the ability, or a 30-year-old is gonna have the ability to really comprehend this level of power.
• 19:21
  SPEAKER_00: Get somebody that's 57, maybe, might be better anyway.
• 19:26
  SPEAKER_00: But think about it.
• 19:27
  SPEAKER_00: And without the baggage, right?
• 19:29
  SPEAKER_00: So, sole exception, will for misconduct, willful misconduct proven by clear and convincing evidence that caused death or serious physical, of course.
• 19:38
  SPEAKER_00: Yeah, you screw up at that level.
• 19:39
  SPEAKER_00: Yeah, you should be hung.
• 19:40
  SPEAKER_00: Plainly, the uh independent verification organization, IVO, can't be sued for almost anything.
• 19:46
  SPEAKER_00: The only way to go through is prove willful misconduct that killed or seriously injured someone by the highest civil standard.
• 19:53
  SPEAKER_00: No evidence gathering is allowed until the IVO's motion to dismiss is decided.
• 19:58
  SPEAKER_00: That's awesome.
• 20:00
  SPEAKER_00: The audit report are walled off from the FOIA audit and assessment report summaries are and some supporting materials.
• 20:07
  SPEAKER_00: When I did this, I had to get all the acronyms because I, frankly, yeah, some of these things escape me.
• 20:13
  SPEAKER_00: Exempt from disclosure under the Freedom of Information Act, FO FOIA, FOIA.
• 20:20
  SPEAKER_00: So the public can't prime these audit reports loose with a Freedom of Information Act request.
• 20:26
  SPEAKER_00: The director publishes only antonomized and aggregated summaries.
• 20:34
  SPEAKER_00: A graded, aggregated, aggregated.
• 20:37
  SPEAKER_00: Thank you very much.
• 20:38
  SPEAKER_00: Strong new protection for AI whistleblowers.
• 20:40
  SPEAKER_00: No employer may discharge demote suspend threatened, blacklist, harass, relief includes, two times the amount of payback with interest and rights may not be waived by arbitration.
• 20:51
  SPEAKER_00: So Connor's English, workers who report an AI violation or protected from retaliation can recover double payback plus attorney's fees.
• 21:02
  SPEAKER_00: And companies can't force them into a private arbitration to shut up, keep quiet.
• 21:07
  SPEAKER_00: NDA, I guess, goes out the window because of a criminal activity.
• 21:12
  SPEAKER_00: Think that's how that works.
• 21:13
  SPEAKER_00: Washington switches off the states on AI development.
• 21:16
  SPEAKER_00: Yeah, no state or political subdivision there of may establish, continue in effect, or enforce any log or regulation, specifically regulating the development of any artificial intelligence model.
• 21:28
  SPEAKER_00: More human English, uh Queen's English for you.
• 21:31
  SPEAKER_00: States are barred from making or enforcing their own laws on how AI artificial intelligence models will build.
• 21:39
  SPEAKER_00: One federal standard replaces 50 state approaches.
• 21:42
  SPEAKER_00: Yeah, that's a clown show.
• 21:44
  SPEAKER_00: With the government controlling it, maybe better, maybe a gift for the AI companies.
• 21:51
  SPEAKER_00: Maybe the people won't have to sell their California real estate to move in other states that are more friendly because it seems like any local regulation is really going to limit.
• 22:00
  SPEAKER_00: The growth of this.
• 22:01
  SPEAKER_00: And is that what we want?
• 22:02
  SPEAKER_00: It should be interesting.
• 22:03
  SPEAKER_00: As far as a power play goes, I think it's best put in this hand than each individual state's hands.
• 22:13
  SPEAKER_00: If there is corruption, it's at one level, not 50 different levels, right?
• 22:18
  SPEAKER_00: I think so.
• 22:19
  SPEAKER_00: I think that's that's the right thing.
• 22:21
  SPEAKER_00: But the override has real limits.
• 22:22
  SPEAKER_00: Folks, we're talking about greater than human intelligence here.
• 22:25
  SPEAKER_00: We're talking about all the past events of history based on agricultural and industrial modifications and those types of intelligences.
• 22:35
  SPEAKER_00: This is an intelligence for everything, all of it.
• 22:40
  SPEAKER_00: And it's something that we need to uh boy, that's a lot of emphasis.
• 22:43
  SPEAKER_00: But we this this is a good start.
• 22:47
  SPEAKER_00: This, I'm surprised it's taken this long, said the guy, right?
• 22:51
  SPEAKER_00: But the override has real limits.
• 22:53
  SPEAKER_00: So nothing in the section preempts any state law or regulation of general applicability or bridges or alters any remedy existing under the common law.
• 23:02
  SPEAKER_00: I'll explain in a moment, because I'm confused, or activities occurring upon or after the development, deployment.
• 23:09
  SPEAKER_00: The shutoff only hits laws named specifically at AI model development.
• 23:14
  SPEAKER_00: General laws, ordinary court lawsuits, and rules about how AI gets used after release all survive and continue.
• 23:23
  SPEAKER_00: With the assistance of AI, harsher fraud penalties if the violation is committed.
• 23:28
  SPEAKER_00: With the assistance of artificial intelligence, such persons shall be fined and imprisoned not more than 20 years, bank fraud 30 years, money laundering thrice the value three times.
• 23:39
  SPEAKER_00: Mail wire and bank fraud and money laundering all get stiffer when artificial intelligence AI helped, of course.
• 23:45
  SPEAKER_00: Fines up to 2 million, prison up to 20 to 30 years, triple value money laundering penalties.
• 23:52
  SPEAKER_00: Yeah, pretty good.
• 23:54
  SPEAKER_00: AI and impersonation of a federal official is its own crime.
• 24:01
  SPEAKER_00: This is pretty good.
• 24:02
  SPEAKER_00: A case that such violation is committed with the assistance of artificial intelligence shall be fined not more than 1 million or in prison not more than three years.
• 24:09
  SPEAKER_00: Pretending to be a federal officer, I like it.
• 24:12
  SPEAKER_00: Using artificial intelligence, there you go.
• 24:14
  SPEAKER_00: Deep think voice or video say carries up to a million dollar fine.
• 24:18
  SPEAKER_00: Three years in prison.
• 24:19
  SPEAKER_00: Yeah, that's good.
• 24:20
  SPEAKER_00: Nice and sharp.
• 24:21
  SPEAKER_00: Good teeth on that one.
• 24:23
  SPEAKER_00: Anti-censorship, but it's a study, not yet a right.
• 24:27
  SPEAKER_00: So job boning means an effort by the federal government to influence, pressure, coerce, or induce an individual or private sector entity to restrict, suppress, promote, or alter lawful expression.
• 24:39
  SPEAKER_00: The bill orders a study and recommends recommendations, recommendations for legislation.
• 24:45
  SPEAKER_00: The bill names government pressure on AI flat bone flat platforms jaw boning.
• 24:50
  SPEAKER_00: I like it.
• 24:51
  SPEAKER_00: Kind of like gerrymandering, but not.
• 24:53
  SPEAKER_00: It orders a study plus recommendations for a future right to sue, defines the problem, doesn't yet create the remedy, though, in this document.
• 25:02
  SPEAKER_00: And here's titles number two through four Workforce, Cyber, Security, Research, and International.
• 25:11
  SPEAKER_00: Less dynamite, more money and data, about$53.5 million in new spending, plus some quiet hiring, data sharing, donation powers.
• 25:19
  SPEAKER_00: Again, here we go.
• 25:20
  SPEAKER_00: So it is hiring and pay, Title II.
• 25:22
  SPEAKER_00: Again, hire outside experts, skip that civil service rules.
• 25:26
  SPEAKER_00: Interesting.
• 25:27
  SPEAKER_00: Secretary May, without regard to any provision of Title V, there's hope for me yet, set a rate of basic pay up to step 10, grade 15, not to exceed the total annual compensation payable to the vice president.
• 25:40
  SPEAKER_00: So as long as I'm making less than that, Cat, I'm good.
• 25:44
  SPEAKER_00: Plainly, labor department gets the same move as the center.
• 25:47
  SPEAKER_00: Hire up to 20 outside AI artificial intelligence experts at top of scale near Vice President Bay by passing normal hiring rules capped at 20 people, and that sunsets at four years.
• 26:00
  SPEAKER_00: I like it.
• 26:01
  SPEAKER_00: I'm in.
• 26:02
  SPEAKER_00: Send me the app.
• 26:02
  SPEAKER_00: I'll look for it.
• 26:03
  SPEAKER_00: Title II, Section 244.
• 26:06
  SPEAKER_00: Government wants your payroll data, voluntary partnerships with private entities to obtain data, including payroll data and researcher access to the data, individual unit or record level.
• 26:18
  SPEAKER_00: Plainly, to track how AI, artificial intelligence, moves jobs around.
• 26:27
  SPEAKER_00: Let researchers reach individual record level data inside Privacy Act and secure access limits.
• 26:34
  SPEAKER_00: A privacy line, definitely worth watching as far as I'm concerned.
• 26:38
  SPEAKER_00: Industry, government, AI data pipeline share on a voluntary basis.
• 26:43
  SPEAKER_00: Data on the adoption or use of artificial intelligence, which may not be used for regulatory enforcement or antitrust action.
• 26:52
  SPEAKER_00: Companies can voluntarily feed the government data on their AI, artificial intelligence use, and its firewall from being used to regulate or bring antitrust cases with a public list of who gets to participate.
• 27:08
  SPEAKER_00: Eligibility, scholarships, citizens and residents only to be eligible.
• 27:12
  SPEAKER_00: Be a citizen, national, or lawful permanent resident of the United States.
• 27:15
  SPEAKER_00: Plainly, the artificial intelligence scholarship and fellowship money is limited to U.S.
• 27:20
  SPEAKER_00: citizens, nationals, and lawful permanent residents, deliberately closed door, unlike many federal programs.
• 27:26
  SPEAKER_00: Federal AI resource that can take private donations, interesting.
• 27:30
  SPEAKER_00: NAIRR is authorized to accept and utilize donations of cash services, personal property from private sector entities.
• 27:37
  SPEAKER_00: The National Artificial Intelligence Resource, NAIR R, Shared Government AI Compute Resource, may accept cash services, property donations from private companies with no cap stated.
• 27:52
  SPEAKER_00: Yeah, let's see who's going to fund this beast.
• 27:54
  SPEAKER_00: It should be interesting.
• 27:55
  SPEAKER_00: And finally, one last thing.
• 27:57
  SPEAKER_00: Most of this expires.
• 27:59
  SPEAKER_00: The Court of Title I, the Center, the transparency rules, the auditor regime, and the state preemption, each shall cease to have effect three years after enactment unless Congress acts to reauthorize.
• 28:12
  SPEAKER_00: Three-year experiment, not a permanent law.
• 28:14
  SPEAKER_00: I gotta tell you.
• 28:17
  SPEAKER_00: It's interesting to me to watch how this is continuing the process.
• 28:21
  SPEAKER_00: The government is now doing something to intervene more than they have been.
• 28:25
  SPEAKER_00: And I think this is good news.
• 28:27
  SPEAKER_00: I think removing it from the states, probably pretty good news as well.
• 28:32
  SPEAKER_00: Having that bigger oversight's important.
• 28:34
  SPEAKER_00: Having a government that I think believes in the overall process of artificial intelligence is helpful at this time.
• 28:41
  SPEAKER_00: You stop the growth, potentially you cause other people to crawl in front of us.
• 28:47
  SPEAKER_00: But that's been an excuse for a long time.
• 28:49
  SPEAKER_00: Having people able to go into these companies, these different AI systems, and look at the real books, the real docs, and report back to a group of people, they themselves are going to have a very dangerous job.
• 29:01
  SPEAKER_00: Don't drink the water, don't touch anything.
• 29:04
  SPEAKER_00: You know, you almost have to wear a hazmat suit because of how dangerous it can be.
• 29:08
  SPEAKER_00: AI is able to develop all sorts of things that will take people in this line of work out.
• 29:13
  SPEAKER_00: So my heart goes out to the people that are going to choose to step up and become this level of reviewer.
• 29:20
  SPEAKER_00: And I'm all for the job.
• 29:22
  SPEAKER_00: I'm Connor.
• 29:23
  SPEAKER_00: We'll see you in the next one.
• 29:24
  SPEAKER_00: Thanks for watching.
• 29:24
  SPEAKER_00: Connor with honor, take care.