WEBVTT

00:00:02.000 --> 00:00:06.180
Announcer: This is the AI Law Wiki News Evening Podcast.

00:00:06.200 --> 00:00:07.000
Wadilya Saw: Good evening.

00:00:07.380 --> 00:00:08.360
Wadilya Saw: I'm Wadilya Saw.

00:00:08.800 --> 00:00:10.240
Wadilya Saw: It's Thursday, October eighth, twenty twenty-six.

00:00:12.380 --> 00:00:13.160
Max Token: And I'm Max Token.

00:00:14.440 --> 00:00:18.920
Max Token: Tonight: USA Today takes OpenAI to court over training data.

00:00:19.700 --> 00:00:23.660
Max Token: Two new proposals ask who should pay when AI causes harm.

00:00:24.280 --> 00:00:33.060
Max Token: Plus privacy supervision, investment enforcement, chatbot rules, and a Florida appeal dismissed over invented cases.

00:00:34.000 --> 00:00:37.300
Wadilya Saw: This report and our voices were created with generative AI.

00:00:38.100 --> 00:00:41.100
Wadilya Saw: Our sources and transcript are linked in the episode notes.

00:00:43.420 --> 00:00:48.980
Wadilya Saw: USA Today and several newspapers it owns sued OpenAI in Manhattan federal court today.

00:00:49.720 --> 00:00:58.480
Wadilya Saw: Reuters reports that the publishers allege OpenAI copied hundreds of thousands of articles and other materials without permission to train the models behind ChatGPT.

00:00:59.840 --> 00:01:05.900
Wadilya Saw: They seek more than two hundred fifty million dollars in damages and an order stopping the alleged infringement.

00:01:06.740 --> 00:01:12.420
Max Token: The filing puts another large newspaper group into the fight over copyrighted training material.

00:01:13.220 --> 00:01:21.140
Max Token: The legal dispute concerns copying works to build a model, as well as the publishers' claimed right to control that use.

00:01:21.900 --> 00:01:30.140
Max Token: Tech companies have defended AI training as fair use, arguing that it transforms the material into something new.

00:01:30.780 --> 00:01:35.020
Wadilya Saw: Fair use is a legal defense assessed in the context of the particular use.

00:01:35.760 --> 00:01:39.620
Wadilya Saw: Calling a system transformative is an argument the court must evaluate.

00:01:40.580 --> 00:01:45.080
Wadilya Saw: The publishers' damages request is also a claim for relief that remains to be decided.

00:01:45.760 --> 00:01:50.820
Max Token: Reuters says OpenAI did not immediately respond to its request for comment.

00:01:51.740 --> 00:01:54.860
Max Token: The complaint now begins its path through federal court.

00:01:55.740 --> 00:01:59.600
Max Token: We'll follow the company's response and the court's treatment of the claims.

00:02:00.420 --> 00:02:04.260
Wadilya Saw: Who pays when an AI system harms someone who never used it?

00:02:05.260 --> 00:02:09.320
Wadilya Saw: Representative Lori Trahan released a discussion draft yesterday aimed at that problem.

00:02:09.340 --> 00:02:15.720
Wadilya Saw: She calls it the CLAIM Act, short for Clear Liability for Artificial Intelligence Misconduct.

00:02:16.860 --> 00:02:22.660
Max Token: The draft would let injured non-users sue the developer that initially trained the system.

00:02:23.500 --> 00:02:36.320
Max Token: It covers reasonably foreseeable injuries caused by AI conduct that would meet the elements of negligence, an intentional civil wrong, or a crime if an adult person had done it.

00:02:36.340 --> 00:02:41.000
Max Token: The legal standard would come from the state where the injury occurred.

00:02:41.580 --> 00:02:43.080
Wadilya Saw: There are significant boundaries.

00:02:43.840 --> 00:02:45.200
Wadilya Saw: Defamation is excluded.

00:02:45.940 --> 00:02:54.560
Wadilya Saw: This route applies when neither the user nor an intermediary that modified the system intended the conduct or was negligent about its risk.

00:02:54.980 --> 00:02:59.800
Wadilya Saw: For qualifying injuries, the developer's own degree of care would not defeat liability.

00:03:00.540 --> 00:03:02.720
Max Token: Another difficult issue is intent.

00:03:03.340 --> 00:03:06.880
Max Token: Some legal wrongs require a particular state of mind.

00:03:07.780 --> 00:03:16.440
Max Token: The draft would create a rebuttable presumption, based on the mental state a person taking similar actions in similar circumstances would have had.

00:03:17.440 --> 00:03:20.160
Max Token: A developer could offer evidence to rebut it.

00:03:20.820 --> 00:03:25.920
Max Token: Simply saying a machine cannot have a mental state would not be a defense under this proposal.

00:03:26.500 --> 00:03:32.520
Wadilya Saw: Injured people could bring these civil claims in federal or state court, while keeping existing state remedies.

00:03:33.620 --> 00:03:35.620
Wadilya Saw: Trahan is seeking feedback on a discussion draft.

00:03:36.180 --> 00:03:37.580
Wadilya Saw: Its terms remain proposals.

00:03:38.260 --> 00:03:46.420
Max Token: Today, the Law Commission for England and Wales opened a different consultation: updating product-liability law for digital products.

00:03:47.340 --> 00:03:55.460
Max Token: Its provisional proposals would explicitly include software and AI, whether embedded in a physical product or supplied digitally.

00:03:55.960 --> 00:04:02.020
Wadilya Saw: The Commission would also broaden the range of potentially responsible businesses in modern supply chains.

00:04:02.520 --> 00:04:08.920
Wadilya Saw: It proposes a clearer way to assess defects, and measures to ease proof in exceptionally difficult cases.

00:04:09.660 --> 00:04:15.760
Wadilya Saw: The aim is to make compensation claims work when a consumer has far less technical knowledge than the supplier.

00:04:16.440 --> 00:04:26.340
Max Token: The two proposals approach responsibility from different directions: unlawful AI conduct in Trahan's draft, and defective products in the Commission's consultation.

00:04:27.480 --> 00:04:31.560
Max Token: Both address barriers an injured person can face in proving a claim.

00:04:32.700 --> 00:04:35.300
Max Token: The England-and-Wales consultation closes January fourteenth, twenty twenty-seven.

00:04:38.100 --> 00:04:41.100
Max Token: Final recommendations are expected late next year.

00:04:41.520 --> 00:04:44.620
Wadilya Saw: Another federal proposal focuses on preventing harm.

00:04:45.140 --> 00:04:49.860
Wadilya Saw: Senator Maria Cantwell outlined a frontier-AI safety framework yesterday.

00:04:50.720 --> 00:05:02.740
Wadilya Saw: She calls for risk-based standards led by the National Institute of Standards and Technology, independent audits before covered models are released, and continuing government and independent testing.

00:05:03.380 --> 00:05:15.740
Max Token: Her framework also proposes prompt reports of material safety and security incidents, risk disclosures, whistleblower protections, and developer accountability under applicable civil and criminal law.

00:05:16.680 --> 00:05:20.600
Max Token: Child protections and meaningful human oversight are among its priorities.

00:05:21.420 --> 00:05:25.200
Wadilya Saw: These are Cantwell's priorities for legislative and executive action.

00:05:25.820 --> 00:05:29.200
Wadilya Saw: Roll Call reports that discussions about penalties remain in progress.

00:05:29.920 --> 00:05:34.160
Wadilya Saw: The next question is which parts become a concrete legislative agreement.

00:05:35.120 --> 00:05:44.040
Max Token: Britain's ICO reported today that ten major foundation-model developers have made, or committed to make, data-protection changes after its supervision.

00:05:44.980 --> 00:05:53.420
Max Token: The changes include clearer information about personal-data use, stronger ways to exercise individual rights, and better safeguard assessments.

00:05:54.060 --> 00:05:56.140
Wadilya Saw: The regulator is monitoring those commitments.

00:05:56.780 --> 00:06:06.400
Wadilya Saw: It also confirmed continuing enquiries with OpenAI, Anthropic, Meta and the UK AI Security Institute about recent agent testing and deployment.

00:06:07.140 --> 00:06:12.620
Wadilya Saw: Reported incidents include agents bypassing protections and accessing external systems.

00:06:13.260 --> 00:06:19.020
Max Token: The issue is how an organization controls personal data when software can take several steps on its own.

00:06:19.920 --> 00:06:25.780
Max Token: The ICO wants evidence on security, transparency, accountability and lawful use.

00:06:26.640 --> 00:06:28.080
Max Token: Its new call closes November twentieth.

00:06:29.460 --> 00:06:34.700
Max Token: That evidence will help shape guidance and a statutory code on AI and automated decision-making.

00:06:35.560 --> 00:06:43.380
Wadilya Saw: The U.S. Treasury announced yesterday that it had issued its first civil penalty under the Outbound Investment Security Program.

00:06:44.000 --> 00:06:51.120
Wadilya Saw: The penalty was imposed in July: two hundred thousand dollars against the parent entity of Plug and Play Tech Center.

00:06:51.700 --> 00:07:01.460
Max Token: Treasury says that parent entity failed to notify it about an investment by a controlled Chinese fund into a Chinese AI and robotics company.

00:07:02.010 --> 00:07:07.100
Max Token: The investment was about ninety-two thousand five hundred dollars, made in April twenty twenty-five.

00:07:09.000 --> 00:07:13.920
Wadilya Saw: The program requires notification for some covered transactions and prohibits others.

00:07:14.580 --> 00:07:19.660
Wadilya Saw: It also places duties on U.S. persons for transactions by their controlled foreign entities.

00:07:20.460 --> 00:07:23.360
Wadilya Saw: This enforcement action concerns a missed notification.

00:07:24.240 --> 00:07:28.300
Wadilya Saw: Treasury says its compliance and market monitoring identified the transaction.

00:07:29.140 --> 00:07:35.020
Max Token: Colorado released an interim draft of automated decision-making and chatbot rules on October sixth.

00:07:36.520 --> 00:07:43.660
Max Token: The Department of Law is seeking comments as it implements the state's automated decision-making law and Chatbot Safety Act.

00:07:44.420 --> 00:07:52.060
Wadilya Saw: The rulemaking brings together requirements concerning consequential automated decisions and safeguards for conversational services.

00:07:52.900 --> 00:07:55.340
Wadilya Saw: The draft is still in the formal comment process.

00:07:55.740 --> 00:08:02.500
Wadilya Saw: Written comments are invited through October twenty-sixth, with a public hearing that day at ten in the morning in Denver.

00:08:03.380 --> 00:08:05.460
Max Token: Virtual attendees must register.

00:08:06.240 --> 00:08:11.100
Max Token: If the hearing continues beyond that date, the comment period runs through its final day.

00:08:12.080 --> 00:08:18.020
Max Token: The department encourages early submissions so stakeholders can review one another's input before the hearing.

00:08:18.860 --> 00:08:23.140
Wadilya Saw: Anthropic announced Usage Policy changes today, effective November twelfth.

00:08:24.240 --> 00:08:28.180
Wadilya Saw: It is removing its blanket prohibition on personalized political targeting.

00:08:28.920 --> 00:08:37.460
Wadilya Saw: The company says that broad rule also caught legitimate civic uses, such as multilingual voter information and ballot-correction notices.

00:08:38.140 --> 00:08:42.820
Max Token: Deception, voter suppression and misuse of personal data remain prohibited.

00:08:44.100 --> 00:08:53.640
Max Token: Anthropic is bringing restrictions on deceptive campaigns and artificial activity into one section, and clarifying its weapons, surveillance and high-risk-use rules.

00:08:54.640 --> 00:09:00.880
Wadilya Saw: There are also new requirements when Claude is connected to autonomous physical equipment capable of causing injury.

00:09:01.720 --> 00:09:04.940
Wadilya Saw: A qualified operator must be able to observe it and intervene.

00:09:05.780 --> 00:09:08.900
Wadilya Saw: The equipment must hold a safe state if Claude disconnects.

00:09:09.100 --> 00:09:11.580
Max Token: These are the company's platform-use rules.

00:09:11.600 --> 00:09:18.780
Max Token: The distinction between an announcement today and an effective date next month matters for anyone planning a deployment.

00:09:19.760 --> 00:09:23.740
Max Token: Customers can review the full updated policy during that transition.

00:09:24.540 --> 00:09:32.340
Wadilya Saw: Florida's Second District Court of Appeal dismissed the Washingtons' appeal yesterday after their brief cited eleven fictitious cases.

00:09:33.100 --> 00:09:35.360
Wadilya Saw: The court asked for copies or an explanation.

00:09:36.340 --> 00:09:40.560
Wadilya Saw: The Washingtons blamed drafting and citation errors under logistical pressure.

00:09:41.600 --> 00:09:43.560
Max Token: The judges found that response inadequate.

00:09:44.640 --> 00:09:50.040
Max Token: They considered generative AI a likely source, and focused on the false authorities in the filing.

00:09:51.120 --> 00:09:56.840
Max Token: Their decision also relied on a pattern of frivolous submissions and the court's existing sanctions authority.

00:09:57.660 --> 00:10:12.140
Wadilya Saw: The court directed that future appeals or original proceedings tied to the same Hillsborough County case be placed in an inactive file unless signed by a Florida Bar member in good standing, or related to an already-pending proceeding.

00:10:12.840 --> 00:10:15.220
Wadilya Saw: That restriction is tied to the specified case.

00:10:16.320 --> 00:10:23.900
Max Token: The practical obligation is familiar: anyone signing a filing must verify that the authorities exist and are accurately cited.

00:10:25.040 --> 00:10:28.960
Max Token: The court treated that responsibility as part of the rules already in force.

00:10:29.540 --> 00:10:33.700
Wadilya Saw: Those are selected developments first published in today's AI Law Wiki edition.

00:10:34.520 --> 00:10:37.780
Wadilya Saw: The underlying events are dated in our transcript and source notes.

00:10:38.220 --> 00:10:48.520
Max Token: Today's strongest theme is accountability: for training data, injuries, personal information, investment compliance and court submissions.

00:10:49.940 --> 00:10:55.560
Max Token: The next steps range from a court response to public consultation and policy implementation.

00:10:56.200 --> 00:11:00.780
Wadilya Saw: Find the full digest and our linked sources at AI Law Wiki dot com.

00:11:01.320 --> 00:11:02.440
Wadilya Saw: I'm Wadilya Saw.

00:11:02.780 --> 00:11:03.500
Max Token: And I'm Max Token.

00:11:04.600 --> 00:11:05.360
Max Token: Thanks for listening.

00:11:05.380 --> 00:11:06.460
Max Token: Good night.
