AI Ethics News — 2026-10-25 (Sun)

Overview

Over the past 48 hours, the ethical questions that surfaced were political bias in chatbots, AI-generated election deepfakes ahead of the US midterms, and the safety of vulnerable users, including teens and people seeking mental health advice. Privacy appeared in a single item on running AI agents locally. (TechCrunch, Tech Policy Press, Mashable, The Independent, TechGig) 【Bias, discrimination and fairness】 - A TechCrunch report says that when forced to pick, Grok's AI model chooses Republicans while OpenAI's picks Democrats. The headline does not give the measurement method or scale. (TechCrunch) - A UN Meetings Coverage and Press Releases item on a Third Committee session carries the headline "'Disability Inclusion Cannot Wait until a Conflict Ends', Third Committee Hears, as Experts Warn of Barriers to Humanitarian Aid, AI Risks". Experts warned of AI risks alongside barriers to humanitarian aid for people with disabilities. (UN Meetings Coverage and Press Releases)

Privacy and surveillance

- TechGig reports that building local AI agents avoids cloud chatbot subscriptions and enhances privacy. This is the outlet's framing; the headline does not say what data is involved or who consents to its use. (TechGig) 【Transparency and accountability】 No notable developments. 【Creators, copyright and consent】 No notable developments.

Disinformation and deepfakes

- Reported: Radio Catskill says AI election misinformation is getting more realistic and that voter vigilance is key. (Radio Catskill) - Reported: KOCO says a Russian disinformation group targeted James Talarico with a baseless claim in an AI-generated video. The headline describes the claim as baseless. (KOCO) - Reported: Tech Policy Press says LLMs are still making election deepfakes just days before the US midterms. (Tech Policy Press) - Countermeasure: Bioengineer.org reports on fighting AI fake news with a five-minute chatbot conversation. The headline does not say how effective this is. (Bioengineer.org) 【Protecting children and vulnerable people】 - Teens: Mashable reports that Meta announced temporary chatbot updates to protect teen users. The company acted; the headline says the updates are temporary. (Mashable) - Chatbot platform: Futurism's headline reads "We Investigated Character.AI and Found a Horror Show. Now It's Bankrupt." Futurism describes its own investigation in these terms, and the headline reports that the company is now bankrupt. (Futurism) - People seeking mental health advice: The Independent reports a psychiatrist's account of four main risks of turning to AI for mental health advice. The headline does not list the four risks. (The Independent) 【Ethical principles and guidelines】 No notable developments.

Japan's AI ethics

No notable developments.

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: what the ethical questions have in common】 - Lack of user control appears in several items. Voters face increasingly realistic AI election misinformation, and teens and mental health users face chatbot risks. In each case the headlines suggest the people affected had little say in how the systems behave. (Radio Catskill, Tech Policy Press, Mashable, The Independent) - Lack of disclosure also recurs. The Grok and OpenAI comparison implies that political leanings are built into models, and the headline does not say how they were set or whether users were told. This is an inference. (TechCrunch) - Possible shifts in decisions: Meta's temporary teen updates and Character.AI's bankruptcy suggest companies may face pressure to act on youth safety. The headlines do not establish a causal link between the two. Users concerned about privacy may weigh local AI agents against cloud services. (Mashable, Futurism, TechGig)

What to watch

Whether election deepfakes continue after the US midterms, and whether the countermeasures reported, such as chatbot conversations, show measurable effect. (Tech Policy Press, Bioengineer.org) 2. Whether Meta's temporary teen chatbot updates become permanent. This is an open question; the headline gives no timeline. (Mashable) 3. Whether the political leanings of AI models, as reported for Grok and OpenAI, draw further testing or disclosure. (TechCrunch)