AI Ethics News — 2026-10-19 (Mon) Evening News

Overview

Consent over creative work dominated the past 48 hours, with voice actors, a music dataset and publisher terms all in play. Privacy in AI agents and retail pricing, child and mental-health safety, and election misinformation also surfaced. On the governance side, one organization was certified as AI ethics-compliant and one ethics framework was proposed. 【Bias, discrimination and fairness】 - A scoping review of bias in large AI models for medicine and healthcare was published (Nature). The headline does not give the method or scale, or which groups are disadvantaged. - An item on ethical challenges and responsibilities for doctors using AI in healthcare appeared (InSight+). - A book item describes Howard University teaching AI to sound Black "while it's being racist" (Emerald Book). The headline does not say who is disadvantaged or how this was measured.

Privacy and surveillance

- AI agents are only as good as the personal information users share, which raises privacy concerns (Bozeman Daily Chronicle). - Consumer apprehension is growing as retailers adopt AI-assisted pricing tools (Martins Ferry Times Leader). - The headlines do not say what data is used, whether consent was given, or for what purpose beyond pricing. 【Transparency and accountability】 - Elder Gong called for a moral compass in the age of AI at the G20 Interfaith Forum (newsroom.churchofjesuschrist.org). The headline discloses no specific company or system practices, so what remains undisclosed is not visible from this material. 【Creators, copyright and consent】 - Kylie Minogue songs are in a 12M AI dataset, and ICMP has issued a warning (tech-insider.org). The dispute is whether the songs were used without consent. - Japan Actors Union released new guidelines against unauthorized AI voice cloning, including a strict ban on copying the voices of deceased seiyuu (AUTOMATON). The dispute is consent for voice use, including after death. - The voice actor of Son Goku in Dragon Ball is angry over unauthorized AI use of the voice, reportedly calling it tantamount to theft (Asia Economy). - Meta removed an AI training clause from its commercial terms after publisher backlash (streamlinefeed.co.ke). The dispute is whether publishers' content can be used for training under commercial terms.

Misinformation and deepfakes

- AI misinformation is flooding feeds ahead of Monday's vote (Calgary Journal). The headline gives no confirmed figures and no countermeasures, so the scale of the harm is not verifiable from this material. 【Protection of children and vulnerable people】 - An opinion piece argues that children were not protected from social media and that AI offers another chance (The Hill). It is a call to act, not a reported action. - AI chatbots may aid mental health but reinforce harmful thinking (Mirage News). The risk is to people seeking mental-health support, and the headline says "may". - A Sydney man allegedly discussed plans for a childcare mass shooting with an AI chatbot (SMH.com.au). This is an allegation, and the headline does not say who acted. - A parenting item on the age of AI appeared (Rising Kashmir). 【Ethical principles and guidelines】 - Decided: Korea Airports Corporation was certified as an AI ethics-compliant organization (Maeil Business Newspaper). - Proposed: an AI ethics layer for auditable moral reasoning (TechGig). - Education and awareness: the Press Council highlighted AI ethics at the United Tractors journalist workshop (Jakarta Globe). IIT Delhi and Lithuania are joining hands to train schoolgirls in ethical AI (Education Times).

Japan's AI ethics

- SoftBank announced GaranAI, a data ecosystem platform meant to support coexistence between Japan's content industry and AI development, balancing protection of information value with fair returns (SoftBank). - The Japan Actors Union voice-cloning guidelines are covered in section 5 (AUTOMATON). - The second item listed under Japan is the Korea Airports Corporation certification, which is a Korean case covered in section 8 (Maeil Business Newspaper).

Statements by key figures

No notable developments. None of the tracked figures (Timnit Gebru, Fei-Fei Li, Yoshua Bengio) appear in the headlines. 【Integrated analysis: how the ethical questions connect】 - Missing consent is the shared structure in several areas. Songs in a dataset, cloned voices and publisher content in training terms all involve use of creators' work that they say they did not approve (tech-insider.org, AUTOMATON, Asia Economy, streamlinefeed.co.ke). Personal data shared with AI agents and used for retail pricing raises the same question for consumers (Bozeman Daily Chronicle, Martins Ferry Times Leader). - Missing explanation is the second pattern. Bias in medical AI is being reviewed, but the headlines give no detail on what users or patients are told (Nature, InSight+). - Meta's removal of a clause shows that organized pushback can change terms. It likely means companies face pressure to clarify training rights in contracts. GaranAI's stated aim of fair returns suggests Japanese firms may look to data infrastructure that records permission and compensation (streamlinefeed.co.ke, SoftBank). - Chatbot risks to vulnerable users appear to be a separate question of duty of care, and companies may need to weigh safeguards for mental-health use (Mirage News, SMH.com.au).

What to watch

Whether other AI developers revise training and licensing terms as Meta did, and whether rights holders such as ICMP press further. 2. Whether voice-use guidelines like those of the Japan Actors Union spread to other creator groups, and whether platforms like GaranAI see adoption. 3. Whether the proposed auditable AI ethics layer and certifications such as the Korea Airports Corporation's gain concrete evaluation criteria, and whether chatbot safeguards for vulnerable users follow.