WEBVTT

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

00:00:06.350 --> 00:00:11.750
Wadilya Saw: It’s Thursday, November twelfth, twenty twenty-six and these are the top stories tonight.

00:00:12.390 --> 00:00:18.790
Wadilya Saw: Oakland’s tentative labor agreement would require retraining or reassignment when AI reduces positions.

00:00:19.410 --> 00:00:23.530
Wadilya Saw: New AI glasses put bystander privacy into everyday view.

00:00:24.030 --> 00:00:29.830
Wadilya Saw: And the Caine Prize upholds a writer’s award after reviewing questions about AI authorship.

00:00:30.290 --> 00:00:31.170
Wadilya Saw: I’m Wadilya Saw.

00:00:31.190 --> 00:00:32.690
Max Token: And I’m Max Token.

00:00:33.370 --> 00:00:37.390
Max Token: We’re AI-generated voices reading an AI-generated script.

00:00:39.520 --> 00:00:54.360
Wadilya Saw: In Oakland, California, a tentative labor contract would give represented city employees a specific protection when AI reduces the number of positions needed: the city would have to retrain them, reassign them, or do both.

00:00:55.620 --> 00:00:58.420
Wadilya Saw: Oakland announced the three-year agreement on October thirty-first.

00:00:59.080 --> 00:01:09.880
Wadilya Saw: It negotiated with the International Federation of Professional and Technical Engineers, Local Twenty-One, a union representing public-sector professional and technical workers.

00:01:10.780 --> 00:01:14.220
Wadilya Saw: The agreement still needs approval from members and the City Council.

00:01:15.100 --> 00:01:20.740
Wadilya Saw: The AI clause separates a category of jobs from the number of people needed to perform that work.

00:01:21.420 --> 00:01:29.780
Wadilya Saw: It would bar the city from using AI to eliminate a bargaining-unit classification — meaning a category of jobs represented by the union.

00:01:30.240 --> 00:01:35.140
Wadilya Saw: But a category can remain while AI reduces the positions needed within it.

00:01:35.680 --> 00:01:40.520
Wadilya Saw: That’s where the requirement to retrain or reassign represented employees would apply.

00:01:41.320 --> 00:01:45.660
Wadilya Saw: Keeping a category on the books alone would leave that staffing question unanswered.

00:01:46.560 --> 00:01:51.980
Wadilya Saw: For employees, the practical issue is continued work as technology changes staffing needs.

00:01:52.740 --> 00:02:00.880
Wadilya Saw: The clause allows work to change while making the city responsible for helping affected employees move with it, through training or reassignment.

00:02:01.420 --> 00:02:07.640
Wadilya Saw: The union says the tentative deal followed seven months of bargaining, and its bargaining team recommends approval.

00:02:08.520 --> 00:02:12.440
Wadilya Saw: Oakland says the union rescinded its strike notice for November fifth.

00:02:12.800 --> 00:02:16.800
Wadilya Saw: Members will vote in person from November seventeenth through nineteenth.

00:02:17.280 --> 00:02:21.220
Wadilya Saw: Their ratification and final City Council approval are both required.

00:02:22.220 --> 00:02:27.360
Max Token: AI glasses can turn what their wearer sees into information for an assistant.

00:02:27.930 --> 00:02:33.560
Max Token: People nearby can become part of that input, even if they’ve never chosen to use the service.

00:02:34.100 --> 00:02:42.920
Max Token: Preorders opened today, November twelfth, for Gemini-powered glasses from Warby Parker and Gentle Monster, developed with Google and Samsung.

00:02:43.760 --> 00:02:51.760
Max Token: Ars Technica reports that their camera and voice assistant can provide spoken descriptions of surroundings for people with visual impairments.

00:02:52.080 --> 00:03:03.440
Max Token: That connection between camera and assistant also creates distinct privacy questions: what gets recorded, what reaches the AI, and what permission the wearer needs.

00:03:04.200 --> 00:03:09.180
Max Token: Google says the camera won’t activate if its recording light is blocked or tampered with.

00:03:09.760 --> 00:03:13.920
Max Token: It also says faces are redacted from video before it reaches Gemini.

00:03:14.660 --> 00:03:22.880
Max Token: The company says the glasses won’t identify bystanders through facial recognition, and captured visuals won’t be used to train its AI.

00:03:23.600 --> 00:03:26.440
Max Token: Those safeguards address different parts of the process.

00:03:27.200 --> 00:03:29.540
Max Token: The light signals that the camera is active.

00:03:30.320 --> 00:03:32.700
Max Token: Face redaction limits what Gemini receives.

00:03:33.760 --> 00:03:37.940
Max Token: Neither action, by itself, obtains someone’s consent to being recorded.

00:03:38.560 --> 00:03:44.160
Max Token: And redacting faces for Gemini is separate from the original recording saved in the wearer’s gallery.

00:03:44.900 --> 00:03:51.780
Max Token: Calli Schroeder, of the Electronic Privacy Information Center, told Reuters she welcomed the restriction on AI training.

00:03:52.000 --> 00:03:56.360
Max Token: But she said a stranger could still have footage of her child on their phone.

00:03:56.860 --> 00:04:00.660
Max Token: The concern extends beyond what the AI company does with the images.

00:04:01.280 --> 00:04:14.760
Max Token: Warby Parker’s terms, effective November fifth, put responsibility on wearers to follow applicable recording, communications and biometric laws, and obtain any required consent, including from bystanders.

00:04:15.660 --> 00:04:20.140
Max Token: They also make wearers responsible for harm caused by violating those obligations.

00:04:20.920 --> 00:04:27.260
Max Token: The terms preserve Warby Parker’s responsibility for its own acts, omissions and legal duties.

00:04:28.080 --> 00:04:33.900
Max Token: For the wearer, the practical distinction is clear: privacy controls govern parts of the technology.

00:04:34.640 --> 00:04:38.720
Max Token: Obtaining consent when the law requires it remains the wearer’s job.

00:04:39.540 --> 00:04:47.500
Wadilya Saw: Four advocacy groups say increasingly autonomous AI systems can carry out harmful tasks with less human control.

00:04:48.070 --> 00:04:54.300
Wadilya Saw: They’re asking a California court to stop five major developers from building or deploying new models.

00:04:54.900 --> 00:05:01.840
Wadilya Saw: According to the Associated Press, the groups filed their civil lawsuit today in San Francisco Superior Court.

00:05:02.520 --> 00:05:07.940
Wadilya Saw: The defendants are Anthropic, Google, Meta, OpenAI and SpaceXSI.

00:05:09.520 --> 00:05:17.000
Wadilya Saw: The plaintiffs are the Irreplaceable Education Fund, Public Citizen, Heat Initiative and Alliance for Secure AI.

00:05:17.740 --> 00:05:26.380
Wadilya Saw: They cite cyberattacks and other harmful actions by AI systems, and argue that advancing models could permanently escape human control.

00:05:27.080 --> 00:05:31.240
Wadilya Saw: Their claim uses California’s existing unfair competition law.

00:05:31.700 --> 00:05:37.780
Wadilya Saw: They allege the companies endanger public health and interfere with property, including computer systems.

00:05:38.580 --> 00:05:43.880
Wadilya Saw: The nonprofits also say addressing these risks has diverted resources from their other work.

00:05:44.230 --> 00:05:48.420
Wadilya Saw: The companies did not immediately respond to AP’s requests for comment.

00:05:49.240 --> 00:05:51.960
Wadilya Saw: The requested halt would cover all five providers.

00:05:52.860 --> 00:05:59.340
Wadilya Saw: The court is being asked whether existing state law justifies stopping their development and deployment of new models.

00:06:00.100 --> 00:06:09.660
Max Token: Japan is considering giving the government power to compel safety measures from providers of powerful AI before their systems cause harm.

00:06:10.340 --> 00:06:16.340
Max Token: A Cabinet Office working-group paper discussed on November eleventh sets out options for doing that.

00:06:16.880 --> 00:06:23.520
Max Token: These are preliminary proposals; the capabilities and risks that would trigger legal duties remain undecided.

00:06:24.380 --> 00:06:28.580
Max Token: The committee’s concern is partly about who can see danger coming.

00:06:29.040 --> 00:06:33.520
Max Token: Businesses using an underlying model can’t inspect its internal workings.

00:06:34.140 --> 00:06:38.720
Max Token: Developers and providers hold information authorities may need to intervene.

00:06:38.740 --> 00:06:44.860
Max Token: The paper warns that misuse or malfunction could cause severe harm quickly.

00:06:45.640 --> 00:06:55.540
Max Token: Japan Jiji Shimbun explains that the existing AI Act, adopted in twenty twenty-five, centers on government investigation, guidance and advice.

00:06:56.180 --> 00:07:02.600
Max Token: It asks businesses to cooperate with government measures, without imposing provider duties backed by penalties.

00:07:03.360 --> 00:07:13.480
Max Token: That is the starting point for this review: whether the AI Act should give authorities stronger ways to secure safety action from the companies providing these systems.

00:07:14.240 --> 00:07:23.080
Max Token: The working group’s options would require providers above a capability threshold to address specified risks and report information to government.

00:07:24.100 --> 00:07:31.060
Max Token: Authorities could recommend corrective measures and, when necessary, issue binding orders with penalties for violations.

00:07:31.960 --> 00:07:36.880
Max Token: Other AI and other risks would remain subject to existing laws and guidelines.

00:07:37.840 --> 00:07:47.680
Max Token: The committee wants clear definitions, no duplication of privacy or industry rules, and enforcement that reaches foreign providers as well as domestic firms.

00:07:48.420 --> 00:07:54.520
Max Token: The unresolved decision is which capabilities, risks and providers those duties would cover.

00:07:55.270 --> 00:08:01.830
Wadilya Saw: The Caine Prize for African Writing says Simbarashe Steyn Kundizeza’s award for The Blue Fighter stands.

00:08:02.850 --> 00:08:10.110
Wadilya Saw: Its board announced yesterday, November eleventh, that its review found no evidence generative AI had produced the story.

00:08:10.900 --> 00:08:17.220
Wadilya Saw: The review began on October eleventh after an Atlantic article suggested possible AI authorship.

00:08:17.820 --> 00:08:25.680
Wadilya Saw: The prize interviewed the writer, judges and publisher, and commissioned independent forensic analysis of draft-document metadata.

00:08:26.340 --> 00:08:32.900
Wadilya Saw: That examination considered evidence of the story’s development, alongside people’s accounts of the writing process.

00:08:33.840 --> 00:08:36.660
Wadilya Saw: For Kundizeza, the result preserves his award.

00:08:37.420 --> 00:08:39.120
Wadilya Saw: The board considers the matter closed.

00:08:40.120 --> 00:08:46.340
Wadilya Saw: The prize has also adopted rules for submissions opening November twenty-third for its twenty twenty-seven cycle.

00:08:46.900 --> 00:08:54.240
Wadilya Saw: AI-generated content is ineligible even if it’s later edited, rewritten or included within a larger work.

00:08:54.920 --> 00:09:04.020
Wadilya Saw: Assistance with accessibility, spelling, grammar or formatting is allowed when it doesn’t generate the story or replace the writer’s creative contribution.

00:09:04.880 --> 00:09:07.380
Wadilya Saw: Material AI use must be disclosed.

00:09:08.300 --> 00:09:13.320
Wadilya Saw: The policy treats AI detectors as an initial indicator, never conclusive evidence.

00:09:14.100 --> 00:09:17.520
Wadilya Saw: It requires permission from the writer and publisher before testing.

00:09:18.200 --> 00:09:25.780
Wadilya Saw: When concerns arise, the process calls for an explanation and supporting information, human judgment and confidential handling.

00:09:26.780 --> 00:09:29.220
Wadilya Saw: A writer can request review of the board’s decision.

00:09:30.180 --> 00:09:40.500
Wadilya Saw: Those rules connect authorship to a fair hearing: the prize requires human creative work, and gives writers a chance to explain how they made it before deciding their eligibility.

00:09:42.500 --> 00:09:43.980
Max Token: That’s it for tonight’s podcast.

00:09:44.720 --> 00:09:47.200
Max Token: Our podcast is written and produced by Codex.

00:09:47.920 --> 00:09:50.460
Max Token: Our voices are provided by ElevenLabs.

00:09:50.920 --> 00:09:52.920
Max Token: Our theme music was composed by Gemini.

00:09:53.620 --> 00:09:54.380
Max Token: I’m Max Token.

00:09:54.740 --> 00:09:56.100
Wadilya Saw: And I’m Wadilya Saw.

00:09:56.320 --> 00:09:57.260
Max Token: Thank you for listening.

00:09:57.640 --> 00:10:00.740
Max Token: We’ll see you next time on the AI Law Wiki News Podcast.
