Latest / DX Today | No-Hype Podcast & News About AI & DX / Democrats Want AI Subpoenas After the Midterms

Transcript

Democrats Want AI Subpoenas After the Midterms · DX Today | No-Hype Podcast & News About AI & DX

• 0:00
  SPEAKER_01: Welcome to the DX Today Podcast, your daily deep dive into the AI ecosystem.
• 0:05
  SPEAKER_01: I'm Brian, and joining me as always is Jessica.
• 0:08
  SPEAKER_00: Thanks, Brian.
• 0:09
  SPEAKER_00: Today we are talking about politics because the midterm results are in for the House.
• 0:13
  SPEAKER_00: And Democrats are already promising a very different posture toward the artificial intelligence industry starting in January.
• 0:20
  SPEAKER_01: Right, and I want to keep this practical.
• 0:22
  SPEAKER_01: Every election produces big promises, so the real question for anyone building or buying AI is what actually changes on a Monday morning once the new majority takes over.
• 0:31
  SPEAKER_00: That is the right frame.
• 0:33
  SPEAKER_00: Let's start with what was said out loud this week.
• 0:35
  SPEAKER_00: Congressman Ted Lew of California, a member of House Democratic leadership, told reporters the new majority would hold AI hearings and pursue what he called reasonable guardrails on AI.
• 0:46
  SPEAKER_01: And he did not hide the contrast.
• 0:48
  SPEAKER_01: He accused the president and congressional Republicans of making a conscious decision to let the AI industry run wild, which he called a bad decision.
• 0:57
  SPEAKER_01: That is campaign language, but it signals where the hearings will point.
• 1:01
  SPEAKER_00: It does, and Lou is not freelancing.
• 1:03
  SPEAKER_00: Minority leader Hakeem Jeffries said weeks before the election that the country is in what he described as an AI safety crisis, and he said he wants bipartisan legislation as soon as Congress returns.
• 1:15
  SPEAKER_01: The detail that caught my attention is the select committee idea.
• 1:19
  SPEAKER_01: Jeffries has been weighing a House Select Committee on AI with subpoena power, and that is a very different tool than a polite hearing where executives read prepared statements.
• 1:28
  SPEAKER_00: Exactly.
• 1:29
  SPEAKER_00: Subpoena power changes the incentives.
• 1:32
  SPEAKER_00: Representatives Bill Foster of Illinois and Ted Liu reportedly pitched the idea to Jeffries this fall, and he has said voluntary industry self-policing is not the right response at this moment.
• 1:43
  SPEAKER_01: As someone who has run real systems, I read that as a warning to every company that has leaned on voluntary safety pledges.
• 1:50
  SPEAKER_01: A pledge is a press release, while a subpoena for internal test results and incident logs is a legal obligation.
• 1:56
  SPEAKER_00: Let me add the political context because it explains the energy here.
• 2:00
  SPEAKER_00: Reporting this week cited an NBC exit poll in which more than half of voters called artificial intelligence and data centers a crucial or major factor in their vote, according to Jim Steyer of Common Sense Media.
• 2:12
  SPEAKER_01: That surprises me a little, honestly, because I assumed pocketbook issues drove most of the vote.
• 2:17
  SPEAKER_01: If AI and data centers were that prominent, then electricity bills and local zoning fights were probably doing a lot of the quiet work.
• 2:25
  SPEAKER_00: I think that is right.
• 2:26
  SPEAKER_00: And it ties to something the privacy professionals at the IAPP wrote this week: that data centers face a growing backlash from voters in both parties, not only one side.
• 2:36
  SPEAKER_01: Which makes the next piece interesting.
• 2:38
  SPEAKER_01: Senator Bernie Sanders and Congressman Greg Casser want to curb what they call superintelligence and place a moratorium on new data centers, yet Jeffries would not commit to their superintelligence ban bill.
• 2:50
  SPEAKER_00: Right.
• 2:51
  SPEAKER_00: So there is already a split inside the majority.
• 2:55
  SPEAKER_00: Moderate Representative Scott Peters of California called the AI discussion pretty populist and pretty devoid of fact, which is a pointed thing for a Democrat to say about his own side's debate.
• 3:06
  SPEAKER_01: I appreciate that candor.
• 3:08
  SPEAKER_01: In my experience, the loudest proposals rarely survive contact with committee staff.
• 3:13
  SPEAKER_01: So I would watch the quieter bills that have bipartisan sponsors, because those are the ones that can actually reach a president's desk.
• 3:20
  SPEAKER_00: Let's name them.
• 3:21
  SPEAKER_00: The Guard Act, led on the Republican side by Senator Josh Hawley, is aimed at protecting children from AI companions.
• 3:29
  SPEAKER_00: And Danny Weiss of Common Sense said it ought to be something Congress does right out of the gate.
• 3:33
  SPEAKER_01: That one makes sense as a first move, since protecting kids from chatbots pulls well and has support across the aisle.
• 3:40
  SPEAKER_01: A quick win early in the session gives the new majority credibility for the harder fights later.
• 3:45
  SPEAKER_00: The second bill is the AI Lead Act from Senators Dick Durbin and Josh Hawley, again bipartisan.
• 3:52
  SPEAKER_00: It would set a liability standard for AI systems and let individuals and state attorneys general sue AI companies when their products cause harm.
• 4:00
  SPEAKER_01: This is the one I would circle if I worked in a general counsel's office.
• 4:03
  SPEAKER_01: Liability changes budgets because once a company can be sued, safety testing stops being a nice to have and becomes a line item with a number attached.
• 4:12
  SPEAKER_00: And the money angle is bigger than people assume.
• 4:15
  SPEAKER_00: Lawyers' price risk, insurers' price risk, and investors' price risk.
• 4:19
  SPEAKER_00: So a credible liability standard could reshape how AI products are priced and who is willing to underwrite them.
• 4:25
  SPEAKER_01: Let me push back a little though.
• 4:27
  SPEAKER_01: Liability rules can also entrench the biggest players, because a giant lab can afford 100 lawyers while a three-person startup cannot, and then regulation accidentally protects incumbents.
• 4:36
  SPEAKER_00: That is a fair concern.
• 4:38
  SPEAKER_00: And it is the argument industry groups tend to make.
• 4:41
  SPEAKER_00: The counterpoint is that clear rules can help small companies too, since they can plan around known standards instead of guessing what a regulator or jury might decide later.
• 4:50
  SPEAKER_01: True.
• 4:51
  SPEAKER_01: A third item on the list is frontier model legislation, which would require independent audits and safety evaluations before a powerful model is released.
• 5:00
  SPEAKER_01: And the reporting notes that some states have already passed early versions of that idea.
• 5:04
  SPEAKER_00: And that brings in the state versus federal tension.
• 5:07
  SPEAKER_00: The IAPP pointed to Colorado's AI Act and noted that Representative Manny Rutinel of Colorado, one of its primary authors, may join the House Energy and Commerce Committee.
• 5:19
  SPEAKER_01: That would be a notable pipeline from state experiments into federal drafting.
• 5:23
  SPEAKER_01: Colorado's law has been a lightning rod for companies, so having its author in the room could mean a lot of lessons learned get written into national language.
• 5:32
  SPEAKER_00: Speaking of energy and commerce, Representative Frank Pollone is in line to chair it, and he has talked about a bold agenda to rein in big tech.
• 5:40
  SPEAKER_00: That committee matters because privacy and AI governance both run through it.
• 5:44
  SPEAKER_01: On the Senate side, the same reporting expects commerce to focus on AI governance, federal privacy, children's online safety, and oversight of the Federal Trade Commission.
• 5:54
  SPEAKER_01: I will say that reports differ on Senate control, so let's stick with what is clear in the House.
• 5:58
  SPEAKER_00: Good call.
• 5:59
  SPEAKER_00: The IAPP also notes that House oversight could dig into how the executive branch itself uses AI, including procurement tied to the government efficiency effort.
• 6:08
  SPEAKER_00: So scrutiny may land on public sector deployments and not only on private labs.
• 6:13
  SPEAKER_01: That is a point people overlook.
• 6:15
  SPEAKER_01: If you sell AI to government agencies, you could end up in an oversight letter next spring.
• 6:21
  SPEAKER_01: And that means documentation, audit trails, and clean contracts suddenly matter much more than a slick demo.
• 6:27
  SPEAKER_00: Now the other side of the ledger.
• 6:28
  SPEAKER_00: The IAPP noted that the administration helped kill AI and privacy bills in Utah and Florida last year.
• 6:35
  SPEAKER_00: And one analyst asked whether the administration will have that same type of power in 2027.
• 6:41
  SPEAKER_01: And that is the real constraint.
• 6:43
  SPEAKER_01: Any bill still needs a presidential signature or a veto-proof majority, so the new house can hold hearings and pass bills, but turning them into law is a much steeper climb.
• 6:53
  SPEAKER_00: Which is why I think the early action is oversight rather than statutes.
• 6:58
  SPEAKER_00: Hearings, document requests, and subpoenas do not need the president's approval, and they can change company behavior simply by making executives explain themselves in public.
• 7:07
  SPEAKER_01: Although a legal expert quoted this week, Omer Tinay of Goodwin said he does not see the election results significantly changing corporate behavior.
• 7:16
  SPEAKER_01: Because economic forces drive what companies do far more than elections do.
• 7:21
  SPEAKER_00: I half agree with him.
• 7:23
  SPEAKER_00: Economics will keep the build-out going, but companies hate surprises.
• 7:27
  SPEAKER_00: And I would bet compliance teams are already updating their risk registers because several large law firms have started hiring to help clients handle congressional scrutiny.
• 7:35
  SPEAKER_01: That hiring detail is telling.
• 7:38
  SPEAKER_01: When law firms add headcount for a specific kind of scrutiny, the market is placing a bet that the scrutiny is real.
• 7:45
  SPEAKER_01: And that bet usually arrives before the first hearing is even scheduled.
• 7:49
  SPEAKER_00: The industry is also thinking about the longer game.
• 7:52
  SPEAKER_00: An open AI lobbyist, Chris Lahaine, said that looking toward 2028 a candidate has to offer a compelling third path between AI without speed limits and control by a self-appointed elite.
• 8:04
  SPEAKER_01: I find that framing revealing.
• 8:06
  SPEAKER_01: It admits that unlimited speed is not politically sustainable while also trying to avoid anything that looks like strict licensing.
• 8:13
  SPEAKER_01: So the fight will be about who defines the middle.
• 8:15
  SPEAKER_00: And voters will not accept an abstract middle.
• 8:18
  SPEAKER_00: They will judge it by results.
• 8:19
  SPEAKER_00: Do their electric bills stay manageable?
• 8:22
  SPEAKER_00: Are their children safe online?
• 8:24
  SPEAKER_00: And can they get a human being when an automated system denies a claim?
• 8:28
  SPEAKER_01: That is the Monday morning test I keep returning to.
• 8:31
  SPEAKER_01: People do not care about frontier model theory.
• 8:34
  SPEAKER_01: They care whether the chatbot their kid uses is safe and whether the data center down the road raises their rates.
• 8:40
  SPEAKER_00: So let's translate for our listeners.
• 8:43
  SPEAKER_00: If you build AI products, expect questions about testing, incident reporting, and child safety.
• 8:48
  SPEAKER_00: And start documenting your evaluation process now because that paper trail will be your best defense.
• 8:53
  SPEAKER_01: If you buy AI products, ask your vendors how they would respond to a liability standard.
• 8:59
  SPEAKER_01: A vendor who has already thought about audits and indemnification is a safer partner than one who changes the subject.
• 9:05
  SPEAKER_00: And if you work in data center development or energy, watch the local politics closely.
• 9:10
  SPEAKER_00: Because bipartisan backlash means community benefit agreements and transparent power sourcing may become the price of getting a project approved.
• 9:18
  SPEAKER_01: I would also tell smaller teams not to panic.
• 9:21
  SPEAKER_01: Rules take time, comment periods are long, and early drafts get rewritten.
• 9:26
  SPEAKER_01: So the smart move is to build good habits rather than to chase every headline.
• 9:31
  SPEAKER_00: One more caution about our own certainty.
• 9:34
  SPEAKER_00: Votes were still being tallied in some races when this reporting came out, and bills announced in November often look very different by the time they pass, if they pass at all.
• 9:43
  SPEAKER_01: Agreed.
• 9:44
  SPEAKER_01: We will keep following the committee assignments, because who chairs what will tell us more than any speech, and the first hearing schedule will show whether this is serious oversight or mostly political theater.
• 9:55
  SPEAKER_00: I would add that a bipartisan start matters.
• 9:58
  SPEAKER_00: If the first moves are child safety and liability bills with Republican TO sponsors, that suggests a path where something real can pass, even under divided power.
• 10:07
  SPEAKER_01: That is where I land too.
• 10:09
  SPEAKER_01: Hearings and subpoenas will make headlines.
• 10:11
  SPEAKER_01: But quiet bipartisan bills on kids and liability are where the practical change is most likely to show up first for working teams.
• 10:18
  SPEAKER_00: Let me connect this to the Federal Trade Commission angle, because Senate Commerce is expected to oversee that agency.
• 10:25
  SPEAKER_00: If the Commission gets more political backing, enforcement against deceptive AI claims and sloppy data practices could become much more active than it has been.
• 10:33
  SPEAKER_01: That matches what I see in practice.
• 10:36
  SPEAKER_01: Most companies are not worried about a sweeping new law.
• 10:39
  SPEAKER_01: They are worried about a regulator asking one pointed question about a product claim they cannot back up with evidence.
• 10:45
  SPEAKER_00: Right.
• 10:46
  SPEAKER_00: And that is why marketing teams should be in this conversation too.
• 10:50
  SPEAKER_00: Overstating what an AI feature can do is a quiet legal risk.
• 10:54
  SPEAKER_00: And a more aggressive oversight climate makes those claims much more expensive to get wrong.
• 10:59
  SPEAKER_01: I would also watch the calendar.
• 11:01
  SPEAKER_01: Congress returns in a few weeks, new members get committee assignments, and leadership sets early priorities, so the first signals about hearings could show up well before the formal start of the new session in January.
• 11:13
  SPEAKER_00: That is a good reminder that the real work happens in staff meetings and drafting sessions, nobody films.
• 11:18
  SPEAKER_00: The bills that look surprising in public were usually shaped months earlier by aides, advocates, and a handful of company lobbyists.
• 11:25
  SPEAKER_01: Which is why I keep saying that engagement beats avoidance.
• 11:28
  SPEAKER_01: Companies that talk to committee staff early, with specific technical details, tend to get more workable rules than companies that only show up after a draft has already been written.
• 11:38
  SPEAKER_00: So the takeaway from me is that the era of purely voluntary AI governance in Washington looks like it is ending.
• 11:45
  SPEAKER_00: And companies that prepare early for scrutiny will be calmer when it arrives.
• 11:49
  SPEAKER_01: And mine is simpler.
• 11:51
  SPEAKER_01: Watch the committees, document your testing, and ask your vendors the hard questions now.
• 11:56
  SPEAKER_01: Because the people who prepare during the quiet period are the ones who are not scrambling later.
• 12:01
  SPEAKER_00: Well said.
• 12:02
  SPEAKER_00: This is a story that will unfold over months, not days, and we will keep bringing you the specifics as hearings get scheduled and bills actually move.
• 12:10
  SPEAKER_01: That's all for today's episode of the DX Today podcast.
• 12:14
  SPEAKER_01: Thanks for listening, and we'll see you next time.