WEBVTT

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

00:00:06.350 --> 00:00:11.970
Wadilya Saw: It's Friday, October twenty-third, twenty twenty-six and these are the top stories tonight.

00:00:12.650 --> 00:00:17.010
Wadilya Saw: Italy's AI decree takes effect, with safeguards for employment decisions.

00:00:17.670 --> 00:00:24.610
Wadilya Saw: A Florida court permits targeted discovery of AI conversations while preserving specific claims of legal protection.

00:00:25.230 --> 00:00:32.050
Wadilya Saw: And Suno acknowledges retaining training recordings as it contests a second copyright lawsuit from major labels.

00:00:32.070 --> 00:00:33.970
Wadilya Saw: I'm Wadilya Saw.

00:00:34.290 --> 00:00:35.570
Max Token: And I'm Max Token.

00:00:36.010 --> 00:00:39.930
Max Token: We're AI-generated voices reading an AI-generated script.

00:00:42.160 --> 00:00:49.840
Wadilya Saw: We begin in Italy, where a national decree implementing parts of the European Union's AI Act entered into force today.

00:00:49.860 --> 00:00:54.680
Wadilya Saw: The official gazette records October twenty-third as the commencement date.

00:00:55.460 --> 00:00:59.480
Wadilya Saw: The decree was issued on October seventh and published the following day.

00:01:00.140 --> 00:01:03.180
Wadilya Saw: Today's milestone is that it has now taken effect.

00:01:03.600 --> 00:01:06.160
Max Token: One provision directly concerns workers.

00:01:06.960 --> 00:01:17.620
Max Token: Article forty-one requires employers using AI in employment decisions to reserve the final decision to a person exercising effective, independent authority.

00:01:18.180 --> 00:01:27.160
Max Token: It covers decisions creating, changing or ending an employment relationship, including disciplinary measures and performance evaluations.

00:01:27.960 --> 00:01:32.080
Wadilya Saw: A human name on the paperwork is not the standard described in the text.

00:01:32.740 --> 00:01:36.980
Wadilya Saw: The person must exercise actual, autonomous decision-making power.

00:01:37.680 --> 00:01:46.160
Wadilya Saw: The provision says the worker's consent does not create an exception, and decisions violating that requirement are void, including dismissals.

00:01:46.180 --> 00:01:48.220
Max Token: There is an important boundary.

00:01:48.960 --> 00:02:00.500
Max Token: The article excludes searching for and selecting candidates from its definition of final decisions establishing employment, even when a candidate does not advance to the next selection stage.

00:02:01.300 --> 00:02:03.960
Max Token: That distinction matters when describing the safeguard.

00:02:04.740 --> 00:02:09.520
Max Token: The text treats final employment decisions and earlier candidate screening differently.

00:02:09.540 --> 00:02:18.480
Wadilya Saw: Workers can also request an understandable explanation of a decision, including AI's contribution and the main parameters considered.

00:02:19.460 --> 00:02:29.720
Wadilya Saw: The decree assigns market oversight to the national cybersecurity agency and specified sector regulators, and provides for complaints through the competent authorities' procedures.

00:02:30.400 --> 00:02:36.380
Max Token: The practical change is a combination of employment safeguards and national oversight machinery.

00:02:37.040 --> 00:02:47.820
Max Token: Applying the human-decision requirement means examining who actually exercises authority over a final decision, alongside the information the worker is entitled to receive.

00:02:48.580 --> 00:02:54.760
Max Token: In Florida, a federal discovery ruling explains when AI conversations may become evidence.

00:02:55.300 --> 00:03:08.540
Max Token: The October twenty-first order in Fontainebleau Florida Hotel's case against Jacob Botach permits requests for nonprivileged AI communications concerning the hotel, its lawyers and its personnel.

00:03:09.220 --> 00:03:15.160
Wadilya Saw: The remaining discovery concerns damages in litigation over allegedly defamatory material.

00:03:15.180 --> 00:03:23.700
Wadilya Saw: The hotel argues that prompts and revisions may reveal intent to harm and help establish its claim for punitive damages.

00:03:24.800 --> 00:03:30.060
Wadilya Saw: The court found a sufficient connection to those disputed issues to allow targeted discovery.

00:03:30.860 --> 00:03:38.120
Max Token: But the judge rejected unrestricted requests covering the litigation generally, court orders and judicial officers.

00:03:38.840 --> 00:03:42.380
Max Token: Those categories reached beyond the permitted damages inquiry.

00:03:42.970 --> 00:03:52.600
Max Token: The order also explains that an otherwise responsive conversation remains within scope if it happens to mention one of those excluded subjects.

00:03:53.020 --> 00:03:56.640
Wadilya Saw: Legal protection must be assessed document by document.

00:03:57.460 --> 00:04:02.440
Wadilya Saw: The court rejected a blanket claim that all the conversations were protected work product.

00:04:03.180 --> 00:04:06.880
Wadilya Saw: That term concerns qualifying materials prepared for litigation.

00:04:07.640 --> 00:04:15.800
Wadilya Saw: It also recognized that particular AI-assisted materials may qualify, depending on why and how they were created.

00:04:16.460 --> 00:04:25.120
Max Token: A party asserting that protection must support the claim and provide a sufficient privilege log identifying what is withheld and why.

00:04:25.690 --> 00:04:29.560
Max Token: AI use alone settles neither access nor protection.

00:04:30.220 --> 00:04:37.400
Max Token: Ordinary discovery rules still require attention to relevance, proportionality and the purpose of the material.

00:04:38.120 --> 00:04:41.240
Wadilya Saw: There is a separate limit on access through the service providers.

00:04:42.100 --> 00:04:45.240
Wadilya Saw: The judge refused to force Botach to consent to disclosure.

00:04:46.160 --> 00:04:51.520
Wadilya Saw: The hotel may pursue narrowed subpoenas, but providers retain their ordinary rights to object.

00:04:51.540 --> 00:04:57.760
Wadilya Saw: The parties must first discuss a production process that accommodates potential work-product claims.

00:04:58.480 --> 00:05:06.760
Max Token: The order therefore opens a defined route to evidence while leaving specific protection claims and provider objections to be resolved.

00:05:07.440 --> 00:05:11.640
Max Token: It does not establish that the conversations prove the hotel's allegations.

00:05:12.400 --> 00:05:19.680
Wadilya Saw: Music Business Worldwide reports that Suno has acknowledged keeping some sound recordings used to train its AI models.

00:05:19.700 --> 00:05:27.540
Wadilya Saw: The acknowledgment appears in its October twentieth answer to Universal and Sony's second Massachusetts copyright lawsuit.

00:05:28.160 --> 00:05:30.480
Wadilya Saw: The publication reported on the filing yesterday.

00:05:31.140 --> 00:05:35.400
Max Token: Suno cites an earlier request from the labels to preserve training data.

00:05:35.920 --> 00:05:40.220
Max Token: It disputes infringement and argues that any copying is fair use.

00:05:40.840 --> 00:05:51.900
Max Token: Retention and lawful use are separate questions: acknowledging that recordings remain available does not resolve whether training or later model development infringed copyright.

00:05:52.600 --> 00:05:59.800
Wadilya Saw: The labels allege that the newer version-six models benefited from outputs of earlier models they say were infringing.

00:06:00.840 --> 00:06:02.180
Wadilya Saw: Suno contests their claims.

00:06:02.860 --> 00:06:07.920
Wadilya Saw: It also says artist names in prompts are converted into descriptions of musical style.

00:06:08.420 --> 00:06:10.640
Max Token: These are positions in a contested lawsuit.

00:06:10.660 --> 00:06:21.520
Max Token: Our account follows the full Music Business Worldwide report; we have not independently reviewed the complete answer or tested the model's handling of artist prompts.

00:06:22.360 --> 00:06:30.600
Max Token: In Brazil, Meta says temporary restrictions on AI-generated or altered content are now operating for the election runoff.

00:06:31.380 --> 00:06:40.260
Max Token: Its October twenty-first update says the period began at midnight yesterday, October twenty-second, and ends at eleven fifty-nine p.m.

00:06:40.680 --> 00:06:44.120
Max Token: on October twenty-sixth, local time in Brazil's capital.

00:06:44.760 --> 00:06:50.700
Wadilya Saw: The company says editing tools can embed industry-standard signals indicating AI use.

00:06:51.420 --> 00:06:58.400
Wadilya Saw: It reads those signals automatically and makes affected posts unavailable in Brazil during the restriction period.

00:06:58.880 --> 00:07:02.840
Wadilya Saw: Meta says the posts become visible in the country again afterward.

00:07:02.860 --> 00:07:08.260
Max Token: This is the company's explanation of how it is implementing electoral requirements.

00:07:09.040 --> 00:07:17.900
Max Token: The mechanism it describes depends on signals attached to content, which can come from ordinary image, video or audio editing tools.

00:07:18.420 --> 00:07:24.320
Wadilya Saw: The immediate consequence for users is temporary loss of visibility in Brazil for affected posts.

00:07:25.260 --> 00:07:30.440
Wadilya Saw: This implementation account is distinct from the separate disputes about equal treatment of campaigns.

00:07:30.460 --> 00:07:34.900
Wadilya Saw: It should not be read as a court finding resolving those disputes.

00:07:36.040 --> 00:07:45.120
Wadilya Saw: Australia's privacy regulator today described inquiries into government agencies' assessment practices, including a department's workplace AI deployment.

00:07:46.100 --> 00:07:55.500
Wadilya Saw: The Office of the Australian Information Commissioner says agencies had considered their obligations and privacy risks, and their assessment records aligned with its guidance.

00:07:55.520 --> 00:08:01.740
Max Token: The findings concern how agencies assess privacy risks and document their decisions.

00:08:02.560 --> 00:08:05.580
Max Token: The regulator did not report an AI-related breach.

00:08:06.500 --> 00:08:16.040
Max Token: The regulator reiterates that the Australian Government Privacy Code requires written privacy impact assessments for projects presenting high privacy risk.

00:08:16.560 --> 00:08:20.940
Wadilya Saw: It encourages early screening to identify projects needing fuller assessment.

00:08:20.960 --> 00:08:31.880
Wadilya Saw: The questions include what personal information is needed, who can access it, whether service providers receive it and how people will be told about its use.

00:08:32.320 --> 00:08:38.580
Max Token: Starting early gives an agency the opportunity to change a design before key decisions are fixed.

00:08:39.320 --> 00:08:44.720
Max Token: Clear records also let it explain to the regulator how it reached its privacy decisions.

00:08:44.740 --> 00:08:51.640
Max Token: The account emphasizes proportionate assessment and privacy built into project planning.

00:08:52.520 --> 00:09:08.200
Max Token: Finally, Reuters reported on Wednesday, October twenty-first, that the political group Leading the Future is spending millions to influence at least forty-eight congressional races while opposing restrictions it considers harmful to AI development.

00:09:08.220 --> 00:09:17.980
Wadilya Saw: The report identifies a twenty-five-million-dollar donation from OpenAI president Greg Brockman and his wife Anna as backing for the group.

00:09:18.500 --> 00:09:23.080
Wadilya Saw: OpenAI says the donation was personal, rather than a company contribution.

00:09:24.080 --> 00:09:27.220
Wadilya Saw: The report does not announce a new donation made today.

00:09:27.700 --> 00:09:32.880
Max Token: Leading the Future says it wants a strong, responsible national AI framework.

00:09:32.900 --> 00:09:43.320
Max Token: Reuters also describes an affiliated nonprofit, Build American AI, advocating a federal standard in place of separate state rules.

00:09:44.040 --> 00:09:50.400
Max Token: Critics question whether that political activity matches industry leaders' public support for safeguards.

00:09:51.080 --> 00:09:56.860
Wadilya Saw: The policy dispute concerns both the strength of regulation and which level of government should set it.

00:09:56.880 --> 00:10:06.700
Wadilya Saw: These are electoral spending and advocacy positions; the report does not establish that Congress has enacted the federal pre-emption its supporters want.

00:10:07.460 --> 00:10:13.120
Wadilya Saw: Source links and fuller summaries are in our show notes and today's October news archive group.

00:10:14.670 --> 00:10:16.140
Max Token: That's it for tonight's podcast.

00:10:16.840 --> 00:10:19.300
Max Token: Our podcast is written and produced by Codex.

00:10:20.020 --> 00:10:22.020
Max Token: Our voices are provided by ElevenLabs.

00:10:23.120 --> 00:10:25.160
Max Token: Our theme music was composed by Gemini.

00:10:25.860 --> 00:10:26.860
Max Token: I'm Max Token.

00:10:26.940 --> 00:10:28.360
Wadilya Saw: And I'm Wadilya Saw.

00:10:28.600 --> 00:10:29.660
Max Token: Thank you for listening.

00:10:29.960 --> 00:10:33.140
Max Token: We'll see you next time on the AI Law Wiki News Podcast.
