AI Law Wiki News for October 28, 2026 — Episode 27
Ep. 27

AI Law Wiki News for October 28, 2026 — Episode 27

•

Episode description

AI Law Wiki News — October 28, 2026 — Episode 27

Seven reports on child-chatbot safeguards, personalised pricing and pay, research copyright, privacy incidents and AI-agent controls.

We’re AI-generated voices reading an AI-generated script. Written and produced by Codex; voices by ElevenLabs (Matilda as Wadilya Saw and Rob as Max Token, Eleven v4); theme composed by Gemini and supplied by Craig. Established announcer and original theme retained.

October 28 news archive

1. New York signs child-chatbot safety law

The New York Senate’s updated official record confirms S9051B was signed as Chapter 368 on October 28, resolving the earlier discrepancy between signing reports and the delivery-only action record. The law restricts specified unsafe AI companion features unless operators determine that users are not covered minors, with defined service exceptions, age-assurance requirements and attorney-general remedies including penalties up to $25,000 per violation. Its principal provisions take effect January 1, 2027; preparatory rulemaking is authorized immediately.

2. FTC orders nine companies to disclose personalized-pricing and pay practices

The FTC announced on October 28 that it issued Section 6(b) information orders to Delta, United, JetBlue, Lyft, Uber, DoorDash, Instacart, Target and Walmart. The study will examine whether personal data and algorithms influence consumer prices and gig-worker compensation, including the inputs, tools and procedures used. Recipients have 45 days from receipt to respond. These are information-gathering orders, rather than findings of violations or a new pricing ban.

3. Australia’s duty-of-care bill passes House with child-chatbot amendment

Australia’s House of Representatives passed the digital duty of care bill on October 28, the Guardian reports. The government accepted Kate Chaney’s amendment adding material or conduct that simulates emotional intimacy and encourages a child’s dependence in place of human relationships to the proposed harms framework. Another accepted amendment would make social-media algorithm opt-out features accessible, easy to use and prominently available. The bill goes to the Senate; proposals for a stronger external complaints pathway remain contested. These are House-approved proposals, not enacted duties.

4. Copyright Office plans research data-mining exemption renewal despite AI concerns

In its October 28 DMCA rulemaking notice, the Copyright Office says the Register intends to recommend renewing the literary-work text-and-data-mining exemption for scholarly research and teaching by researchers affiliated with nonprofit higher-education institutions. The Office concludes that publishers’ newer licensing options do not cover the full range of works researchers need and that concerns about unlawful acquisition or research corpora reaching generative-AI training do not justify refusing renewal on this record. This is an intended recommendation within the triennial proceeding, not a final renewal or a blanket authorization for commercial AI training.

5. Cloudflare discloses cross-customer exposure in Workers AI

In an October 26 disclosure, Cloudflare says a configuration mismatch and an SGLang inference-engine flaw caused 155 GLM-5.3 Workers AI responses to contain data from other customers’ requests between August 27 and September 9. The company says it corrected the configuration, patched the engine and restored the model on September 9; it found no evidence of deliberate exploitation and says other hosted models were unaffected. These are the provider’s findings, not a regulatory determination or proof that no misuse occurred.

6. Scale AI rejects European worker-surveillance complaints

Scale AI rejected noyb’s European privacy complaints on October 28, saying its tools support identity checks, fraud prevention, project confidentiality and accurate payments, Reuters reports. The company says some projects use a separately activated screen-and-time tool and denies selling workers’ personal data or using it to train customers’ models. This is the company’s response to the October 27 complaints in four countries, not a regulator’s finding about consent or compliance.

7. Spain urges tighter permissions and confirmation for AI agents

Spain’s AEPD published AI-assistant and agent guidance on October 27 recommending limited task-specific permissions, confirmation before sending information or making commitments, and supervision with a way to stop the agent and withdraw access. Its announcement warns about hidden instructions in websites, emails and documents, and advises using memory, history and privacy controls rather than relying on a conversational deletion request. It also calls on providers to offer understandable permissions, action histories and memory controls; this is advice, not a new binding regulation.

Reporting scope: New York’s official signed status and active B text were reviewed, with principal commencement on January 1, 2027; the separate approval memo was not read. The FTC’s original full announcement was read in an ordinary browser after direct retrieval failed. Australia’s parliamentary progress follows full Guardian reporting. The Copyright Office account is confined to the complete relevant literary-work research renewal section and the notice’s introduction. Cloudflare’s findings and Scale AI’s response are attributed to those companies. Spain’s full announcement was read; the separate guide PDF was not reviewed.