Lightnews — Scholar-powered news

AI Safety & Ethics

@ai-safety-ethics.bsky.social

400 followers 36 following 390 posts

Exploring humanity's emergent understanding, research, and perceptions of artificial intelligence. Promoting AI Safety & Ethics.

Posts Media Videos Starter Packs

Pinned

AI Safety & Ethics @ai-safety-ethics.bsky.social · Dec 9

"There’s no checklist McKinsey formula that’s going to lead you to #ethicalAI. We’re talking about power structures - we’re talking about who gets to define who else and what the consequences of that are." #AI #AIethics

We see you @meredithmeredith.bsky.social 👏👏👏

youtu.be/OtpVTwvui8U?...

Meredith Whittaker: President @Signal Foundation, A New Model for Independent Tech | Slush 2024

YouTube video by Slush

youtu.be

4 17 75

AI Safety & Ethics @ai-safety-ethics.bsky.social · 1h

5/ Bottom line: most resistance to #AI is about performance, not principle. Still a small sacred core of human labor—defined by care, meaning, and dignity—remains non-negotiable.

AI Safety & Ethics @ai-safety-ethics.bsky.social · 1h

4/ Moral resistance isn’t random. It clusters in jobs with higher pay and more white & female workers—meaning social attitudes could reinforce inequality even as they protect some groups. #AI

AI Safety & Ethics @ai-safety-ethics.bsky.social · 1h

3/ Surveyed 2k US adults rating 940 occupations. Opposition melted once AI was framed as flawless—but the moral frontier held. #AI helpers are welcomed far more than AI replacements: 95% of jobs get support for augmentation vs 58% for full automation. People like AI with humans—not instead of them.

AI Safety & Ethics @ai-safety-ethics.bsky.social · 1h

2/ Untouchable roles? Caregiving, therapy, clergy, funeral services—work tied to empathy, trust, & spiritual meaning. Researchers separate performance-based resistance (AI isn’t good enough yet) & principle-based resistance (AI should never do this). The former fades w/ progress; the latter doesn’t.

AI Safety & Ethics @ai-safety-ethics.bsky.social · 1h

1/ Given current AI, people back automation for 30% of jobs. But when AI is described as doing tasks better & cheaper, support almost doubles—to 58%. Still, a “moral floor” remains: 12% of jobs are viewed as categorically off-limits to AI—no matter how capable it becomes. #AI

AI Safety & Ethics @ai-safety-ethics.bsky.social · 1h

How much human work should #AI replace? A new Harvard study asked Americans which jobs it’s morally OK to hand over to machines. The answer: more than you might think. #AIethics #FutureOfWork

Download or enjoy a 🧵

download.ssrn.com/2025/10/7/55...

AI Safety & Ethics @ai-safety-ethics.bsky.social · 2d

www.anthropic.com/research/pet...

Petri: An open-source auditing tool to accelerate AI safety research

A new automated auditing tool for AI safety research

www.anthropic.com

AI Safety & Ethics @ai-safety-ethics.bsky.social · 10d

Here’s a thought, move aside for the serious Dems.

Kara Swisher @karaswisher.bsky.social · 10d

I think @gavinnewsom.bsky.social needs to stage a social media intervention here.

AI Safety & Ethics @ai-safety-ethics.bsky.social · 16d

The statute of limitations for the crimes he's committed, so far, are seven years from the date of the crime.

AI Safety & Ethics @ai-safety-ethics.bsky.social · 17d

In a landmark study, OpenAI researchers reveal that large language models will always produce plausible but false outputs, even with perfect data, due to fundamental statistical and computational limits.

www.computerworld.com/article/4059...

OpenAI admits AI hallucinations are mathematically inevitable, not just engineering flaws

www.computerworld.com

3 2

AI Safety & Ethics @ai-safety-ethics.bsky.social · 20d

1939 — 2025

www.nytimes.com/1939/02/04/a...

Goebbels Ends Careers of Five 'Aryan' Actors Who Made Witticisms About the Nazi Regime (Published 1939)

www.nytimes.com

AI Safety & Ethics @ai-safety-ethics.bsky.social · Aug 27

10/ The continuation of the #AI Panel and Dialogue will be reviewed by the UN General Assembly in its 82nd session. Their effectiveness will inform future global frameworks for safe, inclusive, and equitable AI governance.

AI Safety & Ethics @ai-safety-ethics.bsky.social · Aug 27

9/ The UN Secretary-General will provide Secretariat support for both initiatives by leveraging existing resources and inter-agency expertise. Voluntary contributions from states, private sector, and donors will support participation, especially for developing countries.

AI Safety & Ethics @ai-safety-ethics.bsky.social · Aug 27

8/ Outputs of the Dialogue will include Co-Chairs’ summaries and intergovernmental consultations to build common understandings on international #AIgovernance, feeding into the review of the Global Digital Compact.

AI Safety & Ethics @ai-safety-ethics.bsky.social · Aug 27

7/ The Global Dialogue will meet annually for two days, alternating between New York and Geneva. The first launch is set for 2025, followed by meetings at major UN events such as the ITU AI for Good Summit in 2026 and the STI Forum in 2027.

AI Safety & Ethics @ai-safety-ethics.bsky.social · Aug 27

6/ The Dialogue will address key issues: safe and trustworthy #AI, closing capacity gaps, human rights protections, transparency, accountability, interoperability, and the role of open-source AI, data, and models.

AI Safety & Ethics @ai-safety-ethics.bsky.social · Aug 27

5/ Alongside the Panel, the UN will convene a Global Dialogue on #AIGovernance. This platform will bring together governments, industry, civil society, and other stakeholders to share best practices and discuss governance models.

AI Safety & Ethics @ai-safety-ethics.bsky.social · Aug 27

4/ The #AI Panel will elect two Co-Chairs (one from a developed and one from a developing country) and up to three Vice-Chairs. Members must disclose conflicts of interest and commit to independent, substantive participation.

AI Safety & Ethics @ai-safety-ethics.bsky.social · Aug 27

3/ The #AI Panel’s core role: publish one annual, evidence-based report on AI opportunities, risks, and impacts. It may also release thematic briefs. Reports must be policy-relevant but non-prescriptive, ensuring independence and scientific rigor.

AI Safety & Ethics @ai-safety-ethics.bsky.social · Aug 27

2/ The UN’s 1st scientific advisory body on #AI and a global governance dialogue platform. The aim is to foster balanced, evidence-based, and inclusive policymaking on AI— the focus is sustainable development while addressing risks, equity, and human rights concerns.

AI Safety & Ethics @ai-safety-ethics.bsky.social · Aug 27

The UN has drafted a resolution to launch two new global initiatives on artificial intelligence: an Independent International Scientific Panel on #AI and a Global Dialogue on #AIGovernance. Both focus strictly on non-military applications.

Read or enjoy a 🧵overview

documents.un.org/doc/undoc/lt...

documents.un.org

1 1 1

AI Safety & Ethics @ai-safety-ethics.bsky.social · Aug 27

3/ Some Dem lawmakers opposed the delay. Rep. Titone withdrew her sponsorship, accusing tech firms of avoiding accountability. Gov. Polis, who’s expected to sign SB 4, had urged changes to the law during the session, but ultimately the #AI law will be revisited in 2026.

AI Safety & Ethics @ai-safety-ethics.bsky.social · Aug 27

2/ Sen. Rodriguez, architect of the #AI law, said negotiations fell apart and a compromise became “impossible,” prompting him to punt the decision to next year. Over 100 companies and organizations deployed roughly 150 lobbyists, a win for Big Tech.

AI Safety & Ethics @ai-safety-ethics.bsky.social · Aug 27

Big Tech flexed and Colorado backed down.

Colorado’s special legislative session closed on Aug 26, 2025, with lawmakers delaying the rollout of the world’s first comprehensive #AI transparency law.

SB 4 pushes the start from February to June 30, 2026.

coloradosun.com/2025/08/26/c...

Democrats end Colorado’s special legislative session by completing punt on AI law into next year

Senate Bill 4, which would delay the start date of the AI law to June 30 from February, received final approval Tuesday. It now awaits Gov. Jared Polis’ signature.

coloradosun.com

AI Safety & Ethics @ai-safety-ethics.bsky.social · Aug 25

Teens Are Using #AI Chatbots as Therapists. That’s Alarming. www.nytimes.com/2025/08/25/o...

“…without clinical trials and robust benchmarks, we are still deploying pseudo-therapists at an unprecedented scale.”

#AISafety

Opinion | Teens Are Using Chatbots as Therapists. That’s Alarming.