WEBVTT

00:00:02.000 --> 00:00:05.440
Announcer: This is the AI Law Wiki News Evening Podcast.

00:00:05.460 --> 00:00:07.020
Wadilya Saw: Good evening.

00:00:07.220 --> 00:00:08.320
Wadilya Saw: It's Monday, October twelfth.

00:00:09.300 --> 00:00:17.240
Wadilya Saw: Britain announces new preparations for serious AI incidents, and European experts recommend stronger oversight of models under development.

00:00:17.980 --> 00:00:25.100
Wadilya Saw: In the United States, a proposed sale of airline employee data brings worker privacy into the debate over AI training.

00:00:25.800 --> 00:00:26.760
Wadilya Saw: I'm Wadilya Saw.

00:00:27.060 --> 00:00:27.900
Max Token: And I'm Max Token.

00:00:28.300 --> 00:00:36.340
Max Token: We'll also cover voice-cloning protections, police data, India's securities regulator and New Zealand election proposals.

00:00:37.280 --> 00:00:40.800
Max Token: This report and the anchor voices were created with generative AI.

00:00:41.620 --> 00:00:43.940
Max Token: Our verified sources are linked in the show notes.

00:00:46.460 --> 00:00:52.660
Wadilya Saw: Britain's government announced today that it will expand pilots using frontier AI to strengthen cyber defences.

00:00:53.020 --> 00:01:01.460
Wadilya Saw: The government says its existing pilots helped identify and fix more than a thousand potential vulnerabilities across government systems.

00:01:02.180 --> 00:01:07.380
Wadilya Saw: It is also developing responses to serious incidents involving autonomous AI agents.

00:01:07.960 --> 00:01:11.720
Max Token: An agent can carry out tasks with limited human supervision.

00:01:12.800 --> 00:01:19.580
Max Token: The government says an incident involving agents could spread faster and more widely than other major emergencies.

00:01:20.460 --> 00:01:24.660
Max Token: That raises a response problem for its existing emergency systems.

00:01:25.540 --> 00:01:31.740
Max Token: The National Cyber Security Centre and AI Security Institute are working on that preparedness.

00:01:32.460 --> 00:01:39.000
Wadilya Saw: Ministers will examine stronger protections through the Cyber Security and Resilience Bill and its wider regime.

00:01:39.740 --> 00:01:43.860
Wadilya Saw: That includes ways to identify and report serious incidents and near misses.

00:01:43.880 --> 00:01:49.460
Wadilya Saw: The details are to be developed with parliamentarians, technical experts and operators.

00:01:50.380 --> 00:01:53.880
Wadilya Saw: These are announced programmes and legislative work still to be shaped.

00:01:54.760 --> 00:02:02.600
Max Token: AI minister Kanishka Narayan urged developers to investigate unexpected behaviour and listen to scientists' warnings.

00:02:03.780 --> 00:02:08.740
Max Token: He said systems should be withheld when developers cannot adequately control the risks.

00:02:09.600 --> 00:02:15.060
Max Token: The official release also says evaluations of the latest advanced models are underway.

00:02:16.180 --> 00:02:22.720
Wadilya Saw: Across the Channel, today's European Commission account supplies the outcome of the scientific-panel meeting held October ninth.

00:02:23.760 --> 00:02:28.180
Wadilya Saw: The AI Office has already requested and received serious-incident reports.

00:02:28.860 --> 00:02:35.780
Wadilya Saw: The Commission says further questions, prepared with the panel, will go to the companies behind recent loss-of-control incidents.

00:02:36.260 --> 00:02:42.660
Max Token: The panel's chair and vice-chair recommend oversight of internal models during development and testing.

00:02:43.540 --> 00:02:52.840
Max Token: They also call for stronger incident-detection and investigation capabilities, international information-sharing and joint crisis response.

00:02:53.880 --> 00:02:58.580
Max Token: A further concern is training methods that make models harder to monitor.

00:02:59.420 --> 00:03:04.160
Max Token: Those are expert recommendations informing the Commission's next steps.

00:03:04.980 --> 00:03:07.560
Wadilya Saw: There's also an industry intervention today.

00:03:08.040 --> 00:03:18.960
Wadilya Saw: The Guardian reports that Microsoft president Brad Smith called for safety obligations towards victims of rogue agents, and emergency brakes that could pause or slow AI.

00:03:19.760 --> 00:03:21.640
Wadilya Saw: Those are Smith's policy proposals.

00:03:22.300 --> 00:03:29.720
Wadilya Saw: Taken together, today's developments concern both preventing harmful behaviour and preparing to respond when safeguards fail.

00:03:30.400 --> 00:03:39.440
Max Token: A proposed ten-million-dollar sale of Spirit Airlines data to Google for AI training is facing worker-privacy objections.

00:03:40.080 --> 00:03:50.620
Max Token: The Electronic Privacy Information Center, or EPIC, announced October second that it and a law professor had filed a supporting brief the previous day.

00:03:51.580 --> 00:03:56.180
Max Token: The brief backs the flight attendants' union's objection in the bankruptcy proceedings.

00:03:56.940 --> 00:03:58.760
Wadilya Saw: That earlier filing is context.

00:03:59.560 --> 00:04:06.340
Wadilya Saw: Today's development is an account by two EPIC lawyers involved in the advocacy, published by Tech Policy Press.

00:04:07.060 --> 00:04:13.180
Wadilya Saw: They describe negotiations over excluding sensitive records and deidentifying other employee data.

00:04:13.640 --> 00:04:17.620
Wadilya Saw: They argue that the proposed protections still leave workers exposed.

00:04:18.420 --> 00:04:25.380
Max Token: According to those lawyers, the data includes payroll, disciplinary information and workplace communications.

00:04:26.600 --> 00:04:33.180
Max Token: They say the original agreement addressed consumer privacy while leaving employee protections unclear.

00:04:34.360 --> 00:04:47.140
Max Token: Their article reports that a recent union filing describes agreement to exclude particularly sensitive categories, including medical and disability information, with further discussions continuing.

00:04:47.520 --> 00:04:53.880
Wadilya Saw: Their central objection is that removing direct identifiers may leave a person identifiable through other details.

00:04:54.620 --> 00:05:00.960
Wadilya Saw: Combining workplace messages and records with outside information can expose identity or sensitive attributes.

00:05:01.820 --> 00:05:06.180
Wadilya Saw: That's the advocates' explanation of why deidentification alone may be insufficient.

00:05:06.920 --> 00:05:10.000
Max Token: They want stronger safeguards and worker control.

00:05:10.880 --> 00:05:15.440
Max Token: The October first brief asks the court to address those privacy concerns.

00:05:16.280 --> 00:05:23.580
Max Token: Their broader policy argument is that employers should have fewer incentives to collect and monetize worker information.

00:05:24.940 --> 00:05:29.700
Max Token: The reported transaction remains proposed, with protections under negotiation.

00:05:30.500 --> 00:05:32.640
Wadilya Saw: The practical issue is a change of purpose.

00:05:33.100 --> 00:05:38.300
Wadilya Saw: Information collected to run a workplace is being proposed for sale as AI training material.

00:05:38.960 --> 00:05:47.100
Wadilya Saw: The dispute brings questions of consent, enforceable promises and employee interests into the treatment of data as a bankruptcy asset.

00:05:47.120 --> 00:05:51.880
Max Token: Lithuania's lawmakers have proposed a legal right to a person's voice.

00:05:52.680 --> 00:06:08.080
Max Token: BNS reporting published by the public broadcaster today says the Civil Code amendments would require consent to reproduce, publish or otherwise use identifiable recordings, imitations and artificially generated voices.

00:06:08.800 --> 00:06:11.860
Wadilya Saw: The proposal includes exceptions for satire and parody.

00:06:12.370 --> 00:06:20.780
Wadilya Saw: Those exceptions depend on avoiding deception about authenticity and protecting interests such as dignity, reputation and privacy.

00:06:21.780 --> 00:06:24.900
Wadilya Saw: Its supporters point to fraud and disinformation risks.

00:06:25.380 --> 00:06:30.180
Wadilya Saw: The government says it will assess existing protections for voices and digital likenesses.

00:06:30.320 --> 00:06:32.580
Max Token: Germany is pursuing another route.

00:06:33.160 --> 00:06:36.680
Max Token: Its cabinet approved a digital-violence draft on October seventh.

00:06:37.820 --> 00:06:46.360
Max Token: The government's account says it would criminalize unauthorized creation and distribution of intimate imagery, including sexual deepfakes.

00:06:48.000 --> 00:06:55.020
Max Token: The proposal would also expand civil tools for victims seeking to identify perpetrators and pursue damages.

00:06:55.600 --> 00:07:00.460
Wadilya Saw: Courts could order platforms or internet providers to supply information needed for those claims.

00:07:00.480 --> 00:07:07.800
Wadilya Saw: Serious rights violations, with further abuse expected, could also support temporary account restrictions.

00:07:08.600 --> 00:07:12.800
Wadilya Saw: The government describes a high threshold intended to protect freedom of expression.

00:07:13.480 --> 00:07:21.880
Wadilya Saw: Both stories concern proposed protections, with Germany at the cabinet-approved draft stage and Lithuania considering Civil Code amendments.

00:07:22.540 --> 00:07:31.280
Max Token: An East Midlands policing unit will stop using Palantir Foundry after a two-year trial, according to BBC reporting today.

00:07:32.380 --> 00:07:42.560
Max Token: The trial began in October twenty twenty-four and used Leicestershire Police data to test AI and machine-learning capabilities for serious-crime investigations.

00:07:43.560 --> 00:07:51.000
Wadilya Saw: The unit says it assessed the project and future funding, then decided against taking it further after Home Office funding ended.

00:07:51.500 --> 00:07:57.580
Wadilya Saw: Palantir must now demonstrate that it has cut links to police systems and deleted all the unit's data.

00:07:57.920 --> 00:08:04.980
Max Token: That requirement concerns the end of the relationship: access must be severed and deletion demonstrated.

00:08:05.940 --> 00:08:08.820
Max Token: The police have not disclosed the precise operational datasets.

00:08:10.520 --> 00:08:17.720
Max Token: Palantir told the BBC that its software had improved services and investigations at other British public bodies.

00:08:18.500 --> 00:08:23.380
Max Token: Today's report records the unit's decision and the assurance it is seeking.

00:08:23.820 --> 00:08:30.840
Wadilya Saw: India's securities regulator says it will shortly issue guidance on responsible AI and machine learning in capital markets.

00:08:31.500 --> 00:08:38.100
Wadilya Saw: Its chairman announced the plan October tenth, according to ANI reporting published today by ETCIO.

00:08:38.720 --> 00:08:49.800
Max Token: The Securities and Exchange Board of India, known as SEBI, envisages a tiered approach with clear accountability, data controls, human oversight and kill switches.

00:08:50.300 --> 00:08:58.580
Max Token: The chairman says regulated firms remain responsible for AI outcomes, data quality, model governance and cyber resilience.

00:08:59.700 --> 00:09:02.040
Wadilya Saw: The announcement describes forthcoming guidelines.

00:09:02.820 --> 00:09:07.080
Wadilya Saw: Market participants still need the actual text to assess specific duties.

00:09:07.920 --> 00:09:16.000
Wadilya Saw: The stated direction is that using AI leaves responsibility with the regulated business, with controls proportionate to the risks.

00:09:16.880 --> 00:09:20.940
Max Token: New Zealand's Green Party announced its AI platform today.

00:09:21.420 --> 00:09:32.520
Max Token: Its Safe, Sovereign AI outline proposes an independent regulator for assurance, audits and education, alongside disclosure and labelling rules.

00:09:33.500 --> 00:09:44.040
Wadilya Saw: The party proposes meaningful human oversight and appeal for decisions affecting rights, welfare, health, education, employment or access to public services.

00:09:44.060 --> 00:09:52.760
Wadilya Saw: It also seeks community control over data, environmental safeguards for infrastructure, and benefits shared with workers and communities.

00:09:53.740 --> 00:09:56.680
Max Token: These are election proposals from the Green Party.

00:09:57.260 --> 00:10:05.220
Max Token: The outline links how AI is governed to who can question a consequential decision, who controls the data and who benefits.

00:10:06.240 --> 00:10:11.540
Max Token: An independent regulator and human appeal are central parts of its proposed answer.

00:10:12.380 --> 00:10:20.640
Wadilya Saw: That's our Monday report: incident preparedness, worker data, deepfake remedies and responsibility for AI use.

00:10:21.320 --> 00:10:27.020
Wadilya Saw: Original sources, today's full digest, the transcript and chapters are in the show notes.

00:10:27.800 --> 00:10:30.180
Max Token: This has been AI Law Wiki News.

00:10:30.900 --> 00:10:31.540
Max Token: Thanks for listening.
