AI Law Wiki News for October 20, 2026 — Episode 21
Ep. 21

AI Law Wiki News for October 20, 2026 — Episode 21

•

Episode description

AI Law Wiki News — October 20, 2026 — Episode 21

AI-generated voices reading an AI-generated script. Wadilya Saw (ElevenLabs Matilda) and Max Token (ElevenLabs Rob), generated with Eleven v4. Written and produced by Codex. Theme music composed by Gemini and supplied by Craig; established announcer retained.

Seven stories from the October 20 New York edition. October 19 decisions, findings and sanctions are dated in the audio; Japan’s later releases remain plans.

October 20 news archive

1. Judge orders Meta to identify who approved shadow-library downloads

TorrentFreak reports that Judge Thomas Hixson ordered Meta on October 19 to identify people who approved acquiring shadow-library datasets and using BitTorrent. The order also covers decisions about seeding or mitigating uploads. The discovery dispute concerns Entrepreneur Media’s copyright claims and related author litigation; liability for the contested torrenting remains unresolved.

2. FCA publishes AI practice findings and calls for reporting significant failures

The Financial Conduct Authority published examples of good practice and weaknesses in firms’ AI use, emphasizing consumer outcomes, accountable senior managers, effective human oversight and resilient supplier arrangements. In an accompanying October 19 speech, Jessica Rusu said major-incident reporting should extend to significant AI failures and near-misses. The practice examples explain application of the existing regulatory framework.

3. Meta pauses Muse conversation sharing after researchers find sensitive chats

Ars Technica reports that Claroc researchers found thousands of discoverable Muse conversation-sharing links and encountered sensitive personal and business information while examining hundreds of chats. Meta told Ars it had paused the sharing feature while reviewing the report and considering changes. The findings raise privacy concerns about how publicly shared AI conversations can be discovered.

4. Australian workplace tribunal’s AI disclosure requirements take effect

From October 20, parties using generative AI to prepare Fair Work Commission documents must explain its use, check accuracy and relevance, and confirm those checks in the document. Witness statements must reflect the witness’s own knowledge and words. Professional representatives must hyperlink cited cases. The Commission warns that noncompliance can affect evidentiary weight, costs or dismissal.

5. Japan begins government AI-agent rollout with governance standards

Japan’s Digital Agency began introducing Agent Chat across ministries as part of an AI-agent environment intended for about 180,000 government employees. It also published Gennai governance, security and safety standards defining responsibilities for users, administrators and developers, developed with the AI Safety Institute. The agency aims to introduce the more advanced Gennai Workshop around December and release open-source versions around February 2027.

6. U.S. sanctions five entities linked to Google trade-secret theft

The State Department designated five China-based entities under the Protecting American Intellectual Property Act on October 19, including Shanghai Zhisuan and venture investor MiraclePlus. Its fact sheet says Linwei Ding marketed Zhisuan as able to replicate Google’s AI infrastructure and that MiraclePlus supplied seed funding for an ownership stake. The designations extend to two MiraclePlus-managed funds and a related enterprise.

7. Spotify reports excluding 50 million slop and spam tracks from recommendations

Spotify says it has excluded more than 50 million slop and spam tracks from recommendations this year as part of its response to AI-generated content, impersonation and deception. Its October 20 update also says artist profile protection, which lets artists review releases naming them, will be available to all artists by year-end. The company acknowledges limitations in AI-detection tools.

Download transcript (.vtt)
0:02

Announcer: This is the AI Law Wiki News Evening Podcast.

0:05

Wadilya Saw: It's Tuesday, October twentieth, twenty twenty-six and these are the top stories tonight.

0:12

Wadilya Saw: A judge wants the names behind Meta's shadow-library downloads.

0:16

Wadilya Saw: Britain's financial regulator spells out what effective AI oversight looks like.

0:21

Wadilya Saw: And Meta pauses a conversation-sharing feature after a new privacy report.

0:27

Wadilya Saw: I'm Wadilya Saw.

0:28

Max Token: And I'm Max Token.

0:30

Max Token: We're AI-generated voices reading an AI-generated script.

0:36

Wadilya Saw: A federal magistrate judge has ordered Meta to identify the people who approved obtaining book collections from shadow libraries, including through BitTorrent.

0:46

Wadilya Saw: The October nineteenth discovery order concerns the copyright dispute involving Entrepreneur Media and related author litigation.

0:54

Wadilya Saw: TorrentFreak reports that Judge Thomas Hixson rejected Meta's narrow account of how those decisions were made.

1:01

Max Token: The distinction is between the employees who operated the downloads and the people who authorized them.

1:08

Max Token: Entrepreneur Media asked for everyone who made, authorized or approved the decisions.

1:14

Max Token: Meta said some early data decisions emerged from team consensus, rather than a formal approval by a single person.

1:21

Max Token: It described decisions by the people running torrents as technical choices.

1:27

Wadilya Saw: The judge said the request covers all decision-makers.

1:30

Wadilya Saw: A collaborative decision still requires identifying the people involved.

1:35

Wadilya Saw: The order also reaches decisions about whether to upload material to other users, or how to limit those uploads.

1:42

Wadilya Saw: That's known as seeding: sharing files while participating in a torrent network.

1:47

Max Token: Those answers matter because Entrepreneur Media says they bear on whether infringement was deliberate and on its distribution and contributory-infringement claims.

1:58

Max Token: Meta says it used measures to avoid seeding and was unaware of anyone who approved distributing the datasets.

2:05

Wadilya Saw: This is an order to provide information in discovery.

2:09

Wadilya Saw: The contested questions about torrenting and distribution remain to be decided.

2:13

Wadilya Saw: The training question and the sharing question have different procedural histories, so a ruling about one doesn't supply an answer to the other.

2:23

Wadilya Saw: The next step here is a fuller account of who made the acquisition and sharing decisions.

2:28

Max Token: In Britain, the Financial Conduct Authority has published detailed examples of good AI practice and weaknesses it has seen in financial firms.

2:37

Max Token: Published October nineteenth and updated today, the findings apply the existing regulatory framework to AI.

2:46

Max Token: They cover consumer outcomes, accountable management and operational resilience.

2:51

Wadilya Saw: A recurring message is that putting a person in the process doesn't automatically make oversight effective.

2:58

Wadilya Saw: Reviewers need the information, expertise and authority to challenge a recommendation.

3:04

Wadilya Saw: In one poor-practice example, human review became a box-ticking exercise: people increasingly accepted AI recommendations, while unclear escalation rules let uncertain outputs proceed.

3:17

Max Token: The better examples make that responsibility concrete.

3:21

Max Token: One firm tied each use of AI to the senior manager responsible for the business area, while a separate owner handled the technical model.

3:30

Max Token: Another tested affordability assessments without showing human reviewers the AI's proposed answer, helping the firm compare independent judgments.

3:40

Wadilya Saw: Suppliers are part of the picture too.

3:43

Wadilya Saw: The regulator describes planning for model changes, keeping inventories, testing vendor assurances and maintaining workable alternatives when a service changes or disappears.

3:53

Wadilya Saw: Its examples are illustrative.

3:57

Wadilya Saw: Firms must judge proportionate controls for their own uses and risks.

4:01

Max Token: In a published October nineteenth speech, Jessica Rusu, the authority's chief data, information and intelligence officer, said major-incident reporting must extend to significant AI failures and near-misses.

4:15

Max Token: She also described work toward a framework for agentic finance, including delegated authority and accountability.

4:24

Max Token: Her comments signal regulatory priorities as systems gain the ability to act for customers.

4:30

Max Token: The practice findings themselves create no new requirements.

4:34

Wadilya Saw: Meta has paused a Muse sharing feature while it investigates how shared conversations can be discovered.

4:41

Wadilya Saw: Ars Technica reports that researchers at the security firm Claroc found sixty-eight hundred conversation-sharing links.

4:49

Wadilya Saw: They examined a few hundred and encountered sensitive personal and business information.

4:55

Max Token: Users had created the links to share particular material.

4:59

Max Token: The concern is that people who were never given a link could discover it through public records associated with its web address.

5:06

Max Token: The researchers described confidential work material, loan information and other private content.

5:14

Max Token: We're leaving identifying details out of this account.

5:17

Wadilya Saw: Meta told Ars it had paused the function while reviewing the report and considering changes to link discovery.

5:24

Wadilya Saw: This is separate from yesterday's report about an agent sending a pickup address.

5:29

Wadilya Saw: Today's development concerns discoverability of shared pages and Meta's response.

5:35

Wadilya Saw: A share link can open a private-looking conversation to a wider audience than its creator expects.

5:41

Max Token: Australia's Fair Work Commission now requires people who use generative AI to prepare case documents to disclose and explain that use.

5:51

Max Token: The requirements apply to all Commission matters from October twentieth.

5:56

Max Token: They cover applications, submissions, witness statements and other documents, including documents edited with AI.

6:03

Wadilya Saw: The person submitting the document must check that its information is accurate, relevant and supported by real evidence and legal authorities, and state that those checks were completed.

6:15

Wadilya Saw: Witnesses must also make sure their statements reflect their own knowledge and words.

6:20

Wadilya Saw: Lawyers and other professional representatives must hyperlink cited cases in documents prepared using generative AI.

6:28

Max Token: The Commission warns that noncompliance can lead to reduced evidentiary weight, costs or dismissal.

6:35

Max Token: It also cautions against putting personal or confidential case information into public AI tools.

6:42

Max Token: The practical responsibility stays with the person presenting the material.

6:47

Wadilya Saw: Japan's Digital Agency is beginning a phased rollout of AI-agent tools across government, accompanied by governance, security and safety standards.

6:58

Wadilya Saw: The environment is intended for about a hundred and eighty thousand government employees.

7:03

Max Token: Agent Chat starts rolling out across ministries today.

7:07

Max Token: A more advanced environment, called Gennai Workshop, is targeted for around December.

7:13

Max Token: The agency plans open-source releases around February twenty twenty-seven, allowing local governments and other organizations to adapt the software.

7:21

Wadilya Saw: The announcement says the standards distinguish the responsibilities of users, administrators and developers, with help from Japan's AI Safety Institute.

7:32

Wadilya Saw: It describes rules for government staff using more capable agents.

7:37

Wadilya Saw: The rollout is beginning in stages, with the later tools and public software releases still planned.

7:43

Max Token: The United States has designated five China-based entities under the Protecting American Intellectual Property Act.

7:50

Max Token: The State Department's October nineteenth fact sheet links the action to theft of U.S. trade secrets and to a business established by Linwei Ding.

8:00

Wadilya Saw: The department says Ding marketed Shanghai Zhisuan as capable of replicating Google's AI infrastructure.

8:08

Wadilya Saw: It says the venture investor MiraclePlus provided seed funding in exchange for an ownership stake.

8:13

Wadilya Saw: The other designated entities are two funds managed by MiraclePlus and a related enterprise.

8:21

Max Token: These are the State Department's determinations supporting sanctions under that statute.

8:26

Max Token: Its fact sheet specifies the applicable statutory sanctions provisions.

8:31

Max Token: The new action reaches the business and funding network; we're not treating it as a new court judgment or inferring an across-the-board asset freeze.

8:40

Wadilya Saw: Spotify says it has excluded more than fifty million slop and spam tracks from recommendations this year.

8:48

Wadilya Saw: That's different from removal from the platform: the company separately reports removing more than seventy-five million spam tracks for serious or repeated rule violations.

8:59

Max Token: In today's update, Spotify says its screening combines listener responses, upload patterns, metadata and outside AI-detection tools.

9:10

Max Token: It acknowledges that detection technology has limits.

9:14

Max Token: The figures are Spotify's account of its own enforcement, rather than an independent measurement of all AI-generated music.

9:22

Wadilya Saw: The company also plans to make artist profile protection available to all artists by year-end.

9:29

Wadilya Saw: That tool lets artists review releases that list them, giving them a way to stop the wrong material appearing on their profiles.

9:38

Wadilya Saw: The update puts content labeling, distribution and impersonation controls alongside the continuing debate over AI and creators' rights.

9:48

Max Token: That's it for tonight's podcast.

9:51

Max Token: Our podcast is written and produced by Codex.

9:54

Max Token: Our voices are provided by ElevenLabs.

9:57

Max Token: Our theme music was composed by Gemini.

10:00

Max Token: I'm Max Token.

10:01

Wadilya Saw: And I'm Wadilya Saw.

10:02

Max Token: Thank you for listening.

10:04

Max Token: We'll see you next time on the AI Law Wiki News Podcast.