WEBVTT

00:00:02.000 --> 00:00:06.270
Announcer: This is the AI Law Wiki News Evening Podcast.

00:00:06.350 --> 00:00:12.050
Wadilya Saw: It's Tuesday, October twenty-seventh, twenty twenty-six and these are the top stories tonight.

00:00:12.730 --> 00:00:16.710
Wadilya Saw: A twenty-five-billion-dollar Facebook privacy judgment in New Mexico.

00:00:17.430 --> 00:00:24.970
Wadilya Saw: British peers advance a plan to investigate AI-agent incidents, while rejecting a direct reporting duty for model developers.

00:00:25.590 --> 00:00:30.510
Wadilya Saw: And new privacy complaints target the data behind AI work and photo editing.

00:00:31.150 --> 00:00:31.950
Wadilya Saw: I'm Wadilya Saw.

00:00:31.970 --> 00:00:33.510
Max Token: And I'm Max Token.

00:00:34.210 --> 00:00:38.050
Max Token: We're AI-generated voices reading an AI-generated script.

00:00:40.230 --> 00:00:45.390
Wadilya Saw: We begin with a major consumer-protection judgment newly covered in today's edition.

00:00:46.090 --> 00:00:57.990
Wadilya Saw: On October twenty-first, a New Mexico judge imposed a twenty-five-billion-dollar civil penalty on Meta over Facebook's public representations about user data and platform practices.

00:00:58.810 --> 00:01:03.310
Wadilya Saw: Judge Francis Mathew reserved the question of an injunction for further briefing.

00:01:03.890 --> 00:01:07.530
Max Token: This is the Facebook privacy case filed in twenty twenty-one.

00:01:08.070 --> 00:01:11.470
Max Token: It is separate from New Mexico's later child-safety case.

00:01:12.050 --> 00:01:18.130
Max Token: The judgment follows a jury's findings of willful violations of the state's Unfair Practices Act.

00:01:18.710 --> 00:01:22.910
Max Token: The representations at issue stretched from twenty ten through twenty twenty-one.

00:01:23.310 --> 00:01:39.310
Wadilya Saw: The findings cover several kinds of promises: how much control users had over their data, access by third-party apps, Facebook's investigation after Cambridge Analytica, and its handling of misinformation, hate speech and equal enforcement.

00:01:40.170 --> 00:01:46.330
Wadilya Saw: The court found a repeated pattern of public assurances that conflicted with internal knowledge and practices.

00:01:46.930 --> 00:01:49.650
Max Token: The size of the penalty needs a little unpacking.

00:01:50.290 --> 00:01:59.710
Max Token: The court calculated more than two hundred nineteen billion dollars using five thousand dollars per unfair-or-deceptive violation found by the jury.

00:02:00.210 --> 00:02:03.850
Max Token: The final decree reduced that amount to twenty-five billion dollars.

00:02:04.750 --> 00:02:08.870
Max Token: It also provides for fifteen percent annual interest until payment.

00:02:09.210 --> 00:02:13.690
Wadilya Saw: So twenty-five billion is the penalty actually imposed in this judgment.

00:02:14.370 --> 00:02:18.630
Wadilya Saw: Further briefing concerns what conduct the court might order Meta to change.

00:02:19.490 --> 00:02:29.250
Wadilya Saw: The ruling's relevance to technology governance is concrete: public claims about data controls and enforcement can carry substantial consumer-protection consequences.

00:02:29.950 --> 00:02:37.370
Wadilya Saw: This judgment addresses Facebook's representations; it doesn't establish a new rule governing every AI system.

00:02:37.970 --> 00:02:46.650
Max Token: In Britain, the House of Lords has agreed changes aimed at learning what happens when oversight or control of an AI agent is lost.

00:02:47.470 --> 00:02:54.490
Max Token: The votes took place yesterday, October twenty-sixth, during Report stage of the Cyber Security and Resilience Bill.

00:02:54.890 --> 00:02:58.410
Wadilya Saw: Peers agreed government amendments seventy and seventy-four.

00:02:59.170 --> 00:03:08.330
Wadilya Saw: Together, they would allow later regulations to confer investigation, enforcement and information-sharing functions on the Secretary of State.

00:03:09.210 --> 00:03:18.690
Wadilya Saw: The trigger would be a security or operational compromise linked to loss of oversight or control of an AI agent, or a risk of such a compromise.

00:03:19.250 --> 00:03:20.590
Max Token: What could that mean in practice?

00:03:21.250 --> 00:03:30.990
Wadilya Saw: The amendment describes preserving and analysing information, working out causes, testing remedies, and sharing lessons to prevent another incident.

00:03:31.750 --> 00:03:37.370
Wadilya Saw: Later regulations could also provide inspection powers and requirements on regulated organisations.

00:03:38.550 --> 00:03:45.670
Wadilya Saw: But the definitions of an AI agent and loss of control, along with the detailed functions, would be set in regulations.

00:03:46.730 --> 00:03:47.930
Wadilya Saw: The bill remains pending.

00:03:47.950 --> 00:03:53.150
Max Token: The sharper dispute was over who must tell the authorities when something goes wrong.

00:03:53.890 --> 00:03:58.290
Max Token: Baroness Harding proposed a direct duty on frontier-model developers.

00:03:59.010 --> 00:04:05.650
Max Token: Her amendment would have required an initial notification within twenty-four hours, followed by fuller reports.

00:04:06.530 --> 00:04:14.130
Max Token: Supporters argued that an affected organisation might not know a model was involved, while the developer held the key information.

00:04:14.150 --> 00:04:21.150
Wadilya Saw: The government preferred to put reporting responsibility on regulated services under the cyber-security framework.

00:04:21.790 --> 00:04:30.510
Wadilya Saw: Minister Baroness Lloyd said regulating frontier-model developers was a different question, and that this bill was not the right framework for it.

00:04:31.110 --> 00:04:38.870
Wadilya Saw: She outlined consultation on agent security requirements, preserving incident information and notifying affected organisations.

00:04:39.950 --> 00:04:44.730
Max Token: The Lords rejected Harding's amendment by one hundred sixty-five votes to ninety-five.

00:04:45.710 --> 00:04:54.710
Max Token: The result is a proposed route to investigation and information gathering, with direct developer reporting left outside this amendment package.

00:04:55.410 --> 00:05:00.110
Max Token: The next details depend on the bill's progress and the promised regulatory consultations.

00:05:00.950 --> 00:05:11.870
Wadilya Saw: In the European Union, privacy group noyb says it filed five complaints today against Scale AI in France, Italy, the Netherlands and Austria.

00:05:12.650 --> 00:05:18.750
Wadilya Saw: The complaints concern Outlier, a platform where freelance workers help produce AI training data.

00:05:19.150 --> 00:05:27.350
Max Token: The group alleges that advertising and analytics trackers send workers' personal data to third parties without a valid legal basis.

00:05:27.370 --> 00:05:37.770
Max Token: It also challenges repeated biometric identity checks involving face video and identity documents, saying the checks collect more data than necessary.

00:05:38.510 --> 00:05:43.210
Max Token: And it says workers' requests for access to their own data went unanswered.

00:05:43.850 --> 00:05:48.550
Wadilya Saw: Noyb is asking data-protection regulators to investigate and consider fines.

00:05:49.330 --> 00:05:55.830
Wadilya Saw: Its announcement says Scale describes the identity checks as a way to prevent fraud and verify workers' skills.

00:05:56.590 --> 00:06:01.070
Wadilya Saw: Reuters reports that Scale did not immediately respond to its request for comment.

00:06:01.610 --> 00:06:04.350
Max Token: These are complaints awaiting regulatory assessment.

00:06:05.190 --> 00:06:15.290
Max Token: They put the privacy of the people training AI under scrutiny, including what a platform collects while they work and what they must surrender to get access to paid tasks.

00:06:16.170 --> 00:06:22.830
Wadilya Saw: Queensland is proposing to criminalise creating non-consensual sexually explicit deepfake images or audio.

00:06:23.710 --> 00:06:29.110
Wadilya Saw: Attorney-General Deb Frecklington introduced the legislation today, according to Australia's ABC.

00:06:29.970 --> 00:06:36.990
Max Token: Frecklington says existing offences target distribution, leaving a gap when harmful material is created.

00:06:37.010 --> 00:06:42.190
Max Token: The proposal would carry a maximum three-year prison term for adult offenders.

00:06:43.010 --> 00:06:46.750
Max Token: ABC reports a maximum of eighteen months for children.

00:06:47.130 --> 00:06:55.230
Wadilya Saw: The bill would also update the definition of an intimate image to cover non-genuine depictions, and add intimate audio material.

00:06:56.090 --> 00:07:00.890
Wadilya Saw: The minister says schools are confronting the harm caused by fabricated sexual imagery.

00:07:00.910 --> 00:07:03.590
Max Token: This is legislation before Parliament.

00:07:04.250 --> 00:07:14.650
Max Token: The distinction between creation and distribution is central: the government wants liability to reach the act of making the material, rather than depend on it being shared.

00:07:15.490 --> 00:07:20.930
Max Token: The proposed penalties and expanded definitions still have to pass through the legislative process.

00:07:21.390 --> 00:07:31.570
Wadilya Saw: A proposed Illinois class action raises a different question about AI images: does editing a face involve collecting a legally protected scan of facial geometry?

00:07:32.690 --> 00:07:39.670
Wadilya Saw: Scott Wilson filed the case on October twelfth in federal court in Chicago against the companies behind Grok.

00:07:39.830 --> 00:07:52.030
Max Token: According to Frankfurt Kurnit's case analysis and today's Top Class Actions report, Wilson alleges that Grok derives facial-geometry scans when people upload faces for analysis or editing.

00:07:52.050 --> 00:08:03.550
Max Token: He says the companies failed to provide the written notice, consent process and public retention policy required by Illinois' Biometric Information Privacy Act.

00:08:03.750 --> 00:08:08.110
Wadilya Saw: The legal theory focuses on what the model allegedly derives from a photograph.

00:08:09.150 --> 00:08:13.910
Wadilya Saw: The law excludes photographs themselves from its definition of a biometric identifier.

00:08:15.210 --> 00:08:18.950
Wadilya Saw: Wilson argues a separate face-specific representation is different.

00:08:18.970 --> 00:08:22.130
Wadilya Saw: He seeks damages and injunctive relief.

00:08:22.790 --> 00:08:25.350
Max Token: Those allegations remain to be adjudicated.

00:08:25.810 --> 00:08:35.030
Max Token: A tool's ability to edit a face is part of the plaintiff's argument; the reports don't establish that a court has accepted his account of the underlying processing.

00:08:35.730 --> 00:08:41.310
Wadilya Saw: New York's court system has released guidance for lawyers preparing papers with generative AI.

00:08:42.270 --> 00:08:43.650
Wadilya Saw: The announcement came yesterday.

00:08:44.160 --> 00:08:50.370
Wadilya Saw: It emphasises independent checks of factual and legal assertions, citations and quotations.

00:08:51.150 --> 00:08:55.550
Max Token: The courts say existing professional responsibilities remain unchanged.

00:08:56.390 --> 00:08:59.270
Max Token: Lawyers remain accountable for the papers they sign.

00:08:59.810 --> 00:09:10.030
Max Token: The announcement describes an informational approach that permits AI use, without a separate AI-use disclosure or certification beyond signing the filing.

00:09:10.050 --> 00:09:22.950
Wadilya Saw: Its practical advice includes checking citations and quotations through authoritative legal research tools, reviewing generated text for bias or inappropriate content, and documenting verification.

00:09:24.230 --> 00:09:27.630
Wadilya Saw: The guide builds on a model rule adopted earlier this year.

00:09:28.370 --> 00:09:33.150
Wadilya Saw: The new development is the guidance explaining those responsibilities in day-to-day drafting.

00:09:33.910 --> 00:09:42.030
Max Token: Finally, the federal government has appealed the California judgment in Anthropic's case against the Department of War and other federal defendants.

00:09:42.870 --> 00:09:47.770
Max Token: The notice was filed yesterday, October twenty-sixth, seeking Ninth Circuit review.

00:09:48.370 --> 00:09:56.630
Wadilya Saw: The two-page filing identifies three decisions: the summary-judgment order, the order of final relief and the final judgment.

00:09:57.330 --> 00:09:59.170
Wadilya Saw: The notice starts appellate review.

00:09:59.190 --> 00:10:03.450
Wadilya Saw: It does not itself reverse those decisions or grant a stay.

00:10:04.130 --> 00:10:09.010
Max Token: This California appeal is separate from the procurement-exclusion proceeding in the D.C.

00:10:09.450 --> 00:10:09.850
Max Token: Circuit.

00:10:10.190 --> 00:10:16.470
Max Token: Keeping those tracks distinct matters when following what relief applies and which court is reviewing it.

00:10:16.990 --> 00:10:22.410
Max Token: The filing we reviewed establishes the appeal, rather than a new appellate ruling on the merits.

00:10:24.190 --> 00:10:25.650
Max Token: That's it for tonight's podcast.

00:10:26.370 --> 00:10:28.830
Max Token: Our podcast is written and produced by Codex.

00:10:29.530 --> 00:10:31.970
Max Token: Our voices are provided by ElevenLabs.

00:10:32.590 --> 00:10:34.690
Max Token: Our theme music was composed by Gemini.

00:10:35.450 --> 00:10:36.210
Max Token: I'm Max Token.

00:10:36.530 --> 00:10:37.870
Wadilya Saw: And I'm Wadilya Saw.

00:10:38.110 --> 00:10:39.110
Max Token: Thank you for listening.

00:10:39.510 --> 00:10:42.670
Max Token: We'll see you next time on the AI Law Wiki News Podcast.
