WEBVTT

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

00:00:06.420 --> 00:00:11.620
Wadilya Saw: The Federal Trade Commission demands safety records from OpenAI and Anthropic.

00:00:11.960 --> 00:00:16.920
Wadilya Saw: Global financial supervisors sharpen their guidance for banks using AI agents.

00:00:17.520 --> 00:00:29.200
Max Token: Also tonight: Apple's trade-secret case, the limits of expert discovery, Denmark's digital-likeness proposal, and election safeguards in Britain and Brazil.

00:00:29.880 --> 00:00:33.080
Max Token: It's Thursday, October fifteenth, twenty twenty-six.

00:00:33.680 --> 00:00:34.680
Max Token: I'm Max Token.

00:00:34.800 --> 00:00:36.300
Wadilya Saw: And I'm Wadilya Saw.

00:00:36.660 --> 00:00:40.240
Wadilya Saw: This script and our voices were created with generative AI.

00:00:40.940 --> 00:00:42.940
Wadilya Saw: This is AI Law Wiki News.

00:00:45.290 --> 00:00:51.300
Wadilya Saw: The Federal Trade Commission has sent civil investigative demands to OpenAI and Anthropic.

00:00:51.320 --> 00:00:54.920
Wadilya Saw: The agency confirmed the subpoenas to CBS News.

00:00:55.800 --> 00:01:02.820
Wadilya Saw: Axios reports that they were delivered on Wednesday, October fourteenth, as part of a consumer-protection investigation.

00:01:03.320 --> 00:01:07.640
Max Token: A civil investigative demand is a compulsory request for evidence.

00:01:08.380 --> 00:01:11.660
Max Token: It can help the agency decide whether to pursue a case.

00:01:12.360 --> 00:01:15.920
Max Token: Receiving one does not establish that a company broke the law.

00:01:15.940 --> 00:01:20.680
Wadilya Saw: According to Axios, each demand runs to more than fifty pages.

00:01:21.460 --> 00:01:28.620
Wadilya Saw: The requested material includes security incidents, safeguarding practices, marketing materials and privacy policies.

00:01:29.580 --> 00:01:35.040
Wadilya Saw: That scope puts both the companies' practices and their public representations under scrutiny.

00:01:35.620 --> 00:01:37.960
Max Token: The consumer-protection connection matters.

00:01:38.760 --> 00:01:47.600
Max Token: If a company tells customers that a product has particular safeguards, investigators can compare those statements with what the company actually does.

00:01:48.460 --> 00:01:54.700
Max Token: The evidence would still have to support any eventual allegation of unfair or deceptive conduct.

00:01:55.260 --> 00:02:01.880
Wadilya Saw: OpenAI told Axios that safety remains a priority and that it will engage constructively with the commission.

00:02:02.620 --> 00:02:06.420
Wadilya Saw: The agency declined to give CBS more details about the investigation.

00:02:07.220 --> 00:02:15.020
Max Token: Axios also reports plans to compel testimony from senior AI executives, citing a person familiar with the matter.

00:02:15.780 --> 00:02:18.620
Max Token: Those reported plans are a next step to watch.

00:02:19.340 --> 00:02:22.700
Max Token: For now, the confirmed action is the demand for records.

00:02:23.500 --> 00:02:27.020
Max Token: There is no finding of liability in the reporting we've reviewed.

00:02:27.540 --> 00:02:34.320
Wadilya Saw: Financial supervisors are also asking how institutions will control increasingly capable AI systems.

00:02:34.960 --> 00:02:43.860
Wadilya Saw: Today, the Financial Stability Board published twelve final practices for responsible AI adoption, following a consultation in June.

00:02:44.600 --> 00:02:48.920
Max Token: The practices cover decisions from the boardroom to the individual application.

00:02:49.700 --> 00:02:53.660
Max Token: Boards and senior managers should set strategy and risk appetite.

00:02:54.380 --> 00:03:04.800
Max Token: Institutions should assign responsibilities, assess risks across a system's life, govern data, evaluate performance and maintain human oversight.

00:03:05.540 --> 00:03:11.340
Wadilya Saw: Cybersecurity and dependence on outside technology providers are part of the same framework.

00:03:12.080 --> 00:03:18.880
Wadilya Saw: The board presents this as a practical toolkit that institutions should adapt to their own circumstances and risks.

00:03:19.720 --> 00:03:23.740
Wadilya Saw: It is guidance, rather than a newly binding worldwide AI law.

00:03:23.760 --> 00:03:30.560
Max Token: Reuters reports a stronger focus in the final version on agents that can plan and take actions.

00:03:31.120 --> 00:03:39.760
Max Token: Higher-risk uses call for more intensive human oversight, with clear, risk-based triggers for when an agent must seek human approval.

00:03:40.400 --> 00:03:45.560
Max Token: The practical question is when a system has to stop and ask a person before proceeding.

00:03:46.240 --> 00:03:54.080
Wadilya Saw: Hong Kong's banking regulator announced its own ART framework today, built around accountability, resilience and trust.

00:03:54.920 --> 00:04:01.940
Wadilya Saw: The Hong Kong Monetary Authority paired that announcement with updated supervisory guidance for responsible AI adoption.

00:04:02.660 --> 00:04:12.980
Max Token: It also announced a training-data sharing platform with Cyberport and the University of Hong Kong, and further sandbox testing of applications that use AI agents.

00:04:13.660 --> 00:04:23.980
Max Token: An upcoming discussion paper will explore accountability tied to agents' intended purposes, a registry of material AI agents and mechanisms for intervention.

00:04:24.680 --> 00:04:27.600
Wadilya Saw: That registry is an idea for the forthcoming paper.

00:04:28.240 --> 00:04:34.920
Wadilya Saw: Banks should distinguish the updated guidance published today from the further supervisory concepts still being explored.

00:04:35.700 --> 00:04:42.700
Max Token: In California, OpenAI has agreed to refrain from using Apple's alleged trade secrets while their lawsuit proceeds.

00:04:43.330 --> 00:04:49.960
Max Token: That's according to an October fourteenth report from 9to5Mac, citing Bloomberg's account of that day's hearing.

00:04:50.560 --> 00:04:57.820
Wadilya Saw: The report says Judge Edward Davila indicated he was highly likely to grant Apple's request for a preliminary injunction.

00:04:58.740 --> 00:05:01.840
Wadilya Saw: His reported indication describes what may come next.

00:05:02.380 --> 00:05:06.400
Wadilya Saw: The article does not establish that the injunction had already been issued.

00:05:06.900 --> 00:05:15.260
Max Token: OpenAI also committed to collecting forensic images from fourteen employees' devices that may contain confidential information.

00:05:16.140 --> 00:05:18.700
Max Token: Those images preserve evidence for examination.

00:05:19.840 --> 00:05:24.740
Max Token: The judge pushed back on a broader request covering repositories and devices.

00:05:25.420 --> 00:05:30.420
Wadilya Saw: The immediate development is the voluntary undertaking and evidence-preservation commitment.

00:05:31.000 --> 00:05:36.320
Wadilya Saw: Apple's underlying trade-secret claims remain allegations to be resolved in the case.

00:05:37.060 --> 00:05:44.580
Wadilya Saw: We're covering the hearing first reported in today's wiki edition, rather than treating the lawsuit's earlier accusations as a new event.

00:05:45.000 --> 00:05:52.240
Wadilya Saw: A newly published account of a Nevada discovery ruling draws a line through an expert's ChatGPT history.

00:05:52.960 --> 00:05:56.240
Wadilya Saw: The ruling itself dates to September twenty-ninth.

00:05:56.260 --> 00:05:59.240
Wadilya Saw: Hunton Andrews Kurth reported the details today.

00:05:59.580 --> 00:06:08.080
Max Token: In Villanueva against the Las Vegas Metropolitan Police Department, an expert used ChatGPT for research and report preparation.

00:06:09.060 --> 00:06:16.620
Max Token: According to the firm's account, the district court rejected an order requiring wholesale production of the expert's AI log.

00:06:16.960 --> 00:06:22.940
Wadilya Saw: Federal discovery rules require disclosure of facts or data an expert considered in forming opinions.

00:06:23.740 --> 00:06:27.020
Wadilya Saw: That can include material considered and then set aside.

00:06:27.680 --> 00:06:36.040
Wadilya Saw: Here, AI-generated case summaries had to be produced, allowing the opposing party to examine them for possible fabricated authorities.

00:06:36.720 --> 00:06:42.380
Max Token: The account does not say the court found that those summaries actually contained fabricated cases.

00:06:43.080 --> 00:06:46.900
Max Token: It says the other side was entitled to investigate that possibility.

00:06:47.700 --> 00:06:55.500
Wadilya Saw: Prompts revealing mental impressions and research strategy, deposition preparation and draft-report material received protection.

00:06:56.380 --> 00:07:00.360
Wadilya Saw: The distinction turns on what the material is and how it was used.

00:07:01.220 --> 00:07:07.380
Wadilya Saw: This ruling offers no blanket promise that everything typed into an AI service is protected from discovery.

00:07:08.060 --> 00:07:16.140
Max Token: Denmark is proposing protection against realistic digital imitations of people's appearance and voices shared without consent.

00:07:16.720 --> 00:07:21.200
Max Token: The culture ministry announced the copyright amendment bill on October eighth.

00:07:21.520 --> 00:07:23.340
Max Token: It entered our news coverage today.

00:07:23.780 --> 00:07:27.840
Wadilya Saw: The proposal covers ordinary people and adds protection for performers.

00:07:28.580 --> 00:07:35.500
Wadilya Saw: It would preserve room for caricature, parody, satire and criticism of those in power or of society.

00:07:36.060 --> 00:07:45.100
Max Token: There are limits to those exceptions where misinformation creates a concrete, serious danger to rights, essential interests, life or property.

00:07:45.700 --> 00:07:51.980
Max Token: That makes the boundary between protected expression and harmful imitation a central part of the proposal.

00:07:52.540 --> 00:07:56.420
Wadilya Saw: The culture minister hopes the measure can take effect at the turn of the year.

00:07:57.060 --> 00:08:01.380
Wadilya Saw: That is the minister's aim for a bill still going through the legislative process.

00:08:02.020 --> 00:08:06.380
Wadilya Saw: Listeners should not treat it as an enacted protection with a settled start date.

00:08:07.040 --> 00:08:12.420
Wadilya Saw: Britain's Electoral Commission is developing rapid alerts about significant election deepfakes.

00:08:12.440 --> 00:08:19.980
Wadilya Saw: Tech Policy Press reports today that the target is to notify candidates and parties within three hours of posting.

00:08:20.520 --> 00:08:26.260
Max Token: The commission's chief executive described a May pilot that analysed about three million posts.

00:08:27.040 --> 00:08:29.840
Max Token: It flagged one hundred forty-seven for human review.

00:08:30.780 --> 00:08:34.160
Max Token: Ten misleading items were referred to platforms or police.

00:08:35.020 --> 00:08:40.540
Max Token: Those are different stages of review, not one hundred forty-seven confirmed falsehoods.

00:08:40.560 --> 00:08:48.100
Wadilya Saw: The focus is material undermining election administration, such as a false claim that polling has been cancelled.

00:08:48.900 --> 00:08:53.900
Wadilya Saw: The three-hour figure is a development target, rather than a guarantee for every post.

00:08:54.540 --> 00:09:05.740
Max Token: In Brazil, the Federal Public Prosecutor's Office opened an administrative investigation on October fourteenth into reported election-period restrictions on Meta and X.

00:09:06.220 --> 00:09:14.920
Max Token: It asked the companies to explain restrictions within five days, including whether they followed court orders or algorithmic decisions.

00:09:15.500 --> 00:09:23.700
Wadilya Saw: Separately, Lula's campaign petitioned the electoral court over alleged unequal restrictions on candidate images in Meta AI.

00:09:24.580 --> 00:09:32.980
Wadilya Saw: Agência Brasil reports Meta's response: Brazilian election rules led it to temporarily prohibit candidate recommendations and images.

00:09:33.000 --> 00:09:38.540
Max Token: The prosecutor's inquiry and the campaign's court petition are separate proceedings.

00:09:39.480 --> 00:09:42.360
Max Token: Claims of unfair treatment remain allegations.

00:09:43.340 --> 00:09:52.240
Max Token: Together, these stories show the challenge of responding quickly to harmful election content while making platform restrictions open to scrutiny.

00:09:53.060 --> 00:09:57.440
Wadilya Saw: Source links, the transcript and chapters are available with this episode.

00:09:58.060 --> 00:10:05.720
Wadilya Saw: Today's selection comes from the October fifteenth AI Law Wiki news archive, with earlier event dates identified in the report.

00:10:06.360 --> 00:10:09.280
Max Token: For AI Law Wiki News, I'm Max Token.

00:10:09.440 --> 00:10:10.740
Wadilya Saw: And I'm Wadilya Saw.

00:10:11.180 --> 00:10:12.100
Wadilya Saw: Thank you for listening.
