Sign in

Anna Neumann

@annaneumann.bsky.social
118 followers 141 following 23 posts

PhD Candidate in Responsible AI @ RC Trust ai ethics, accountability, fairness and policy feminist cs

PostsRepliesMedia
Reposted by Anna Neumann
Tech Policy Press @techpolicypress.bsky.social · 11/08/2026
Anna Neumann, Holli Sargeant and Jat Singh argue that system prompts alone don't predict model behavior, so AI safety assessments must evaluate a system's outputs, not solely its written instructions.
techpolicy.press
Making AI Safe Requires More Than Constraining Its Instructions
System prompts alone don't predict model behavior, so "prompt governance" is insufficient for AI safety, write Anna Neumann, Holli Sargeant and Jat Singh.
0123
Anna Neumann @annaneumann.bsky.social · 11/08/2026
Our piece on prompt governance and its insufficiency for effective AI governance is out on @techpolicypress.bsky.social! "Making AI Safe Requires More Than Constraining Its Instructions" argues for better governance of system prompts/instructions: techpolicy.press/making-ai-s/
techpolicy.press
000
Reposted by Anna Neumann
Holli Sargeant @hollisargeant.bsky.social · 11/08/2026
Policymakers increasingly assume inspecting an AI’s system prompts proves it's safe. In @techpolicypress.bsky.social, we show system prompts alone don't predict model behaviour. AI safety assessments must evaluate system outputs, not solely written instructions. w/ @annaneumann.bsky.social
techpolicy.press
Making AI Safe Requires More Than Constraining Its Instructions
System prompts alone don't predict model behavior, so "prompt governance" is insufficient for AI safety, write Anna Neumann, Holli Sargeant and Jat Singh.
021
Reposted by Anna Neumann
Holli Sargeant @hollisargeant.bsky.social · 27/06/2026
@annaneumann.bsky.social just gave a fantastic presentation of our #FAccT2026 paper, "Prompt Governance? On Governing Technologies Governed by Natural Language." Read it here: doi.org/10.1145/3805...
081
Anna Neumann @annaneumann.bsky.social · 09/06/2026
Reading AI slop about your work is one thing Reading viral and *wrong* AI slop that totally misrepresents your paper is another thing that is a shockingly jarring experience
100
Reposted by Anna Neumann
Holli Sargeant @hollisargeant.bsky.social · 09/06/2026
Bumping our paper as we get closer to #FAccT2026! After a 41-day wait, our paper finally made it through the arXiv machinery! If you're interested in whether system prompts serve as robust governance, have a read and come say hi in Montréal. 🔗 arxiv.org/abs/2606.075... @annaneumann.bsky.social
arxiv.org
041
Reposted by Anna Neumann
Kevin Schaul @kevinschaul.bsky.social · 11/05/2026
New from me: See the hidden rules behind AI. Then use them to rewrite this article. I hooked this article up to an LLM to help explain system prompts. Give it a try -> www.washingtonpost.com/technology/i...
AI experiment

Adjust the system prompt to see how an AI system rewrites the article’s introduction.

User prompt
Rewrite this article: Chatbots like ChatGPT are powerful because of their simplicity: Ask just about...

System prompt
You are an AI Experiment by The Washington Post. Always ensure your response uses lots of emojis / mentions goblins / is written for 5-year-olds
1107
Reposted by Anna Neumann
Timnit Gebru @timnitgebru.blacksky.app · 19/02/2026
I'm gonna repeat my comment. Talking about nonexistent "super intelligent" machines is like talking about the Cookie Monster rather than corporations creating real things causing harm. A non existent machine god bringing apocalypse or utopia is a framing that is harmful. Period.
5940241
Reposted by Anna Neumann
tante @tante.cc · 19/02/2026
"Hating technology" in their reading means "not just embracing its inevitability". Maybe that's the reason: Leftist thought is based on the understanding that politics is a common project that allows us to make decisions about the world. The idea of "inevitable technology" is the antithesis to that
114130
Reposted by Anna Neumann
Ketan Joshi @ketanjoshi.co · 08/11/2025
When someone builds a cheap and ubiquitous slop firehose, we're all forced to close our windows and lock our doors Still waiting for someone to explain how this has been a net benefit for human society www.404media.co/arxiv-change...
404media.co
arXiv Changes Rules After Getting Spammed With AI-Generated 'Research' Papers
Cornell University’s arXiv will no longer accept Computer Science reviews and position papers.
615148
Reposted by Anna Neumann
Eryk Salvaggio @eryk.bsky.social · 07/11/2025
There's a lot of confusion about saying all LLM behavior is "learned." Plenty is externally designed, albeit in different ways than we would design any other product. Developers can steer models through data curation, fine tuning, system prompts, RLHF and other directed RL techniques, and so on...
1328
Anna Neumann @annaneumann.bsky.social · 31/07/2025
Launched my Substack! ✨📌 First post is about Trump's AI executive order requiring system prompt disclosure to prove AI isn't 'woke.' It's not making AI neutral, just embedding Trump's ideology while claiming objectivity. Check it out: researchagainstthemachine.substack.com/p/donald-tru...
researchagainstthemachine.substack.com
Donald Trump and the 'Woke' System Prompt Police
Why Trump's system prompt disclosure won't make AI 'unbiased', and might make it worse
020
Reposted by Anna Neumann
Eryk Salvaggio @eryk.bsky.social · 25/07/2025
There could never be a 'Woke AI.' On a technical level, these are machines trained to identify and reproduce statistical biases in uncurated datasets. They are constrained only by the instructions written by the companies building them. The order just demands AI companies do it in a specific way. ⬇️
techpolicy.press
AI Could Never Be ‘Woke’ | TechPolicy.Press
The Trump administration’s plan, in targeting “ideological bias” and “social engineering agendas” in AI, ultimately enforces them, writes Eryk Salvaggio.
43814
Anna Neumann @annaneumann.bsky.social · 25/07/2025
Anthropic: "omg guys, our model can't even keep an in-house vending machine profitable, lol" Also Anthropic: "oh but please use it for your high-stakes financial decisions 🥺"
010
Reposted by Anna Neumann
Tech Policy Press @techpolicypress.bsky.social · 26/06/2025
Data & Society’s Jacob Metcalf writes that whenever you hear someone advocating for acceleration, you should ask: accelerating toward what? Today’s techno-accelerationists aim to undo generations of slow, arduous progress toward social, political, and economic equality, he says.
techpolicy.press
Trump 2.0 Runs on Tech Accelerationism | TechPolicy.Press
Jacob Metcalf says the alliance between the American far right and the tech oligarchy is driven by a shared commitment to accelerationism.
2136
Reposted by Anna Neumann
Melanie Walsh @mellymeldubs.bsky.social · 26/06/2025
Ohhhh Molly Crockett‘s @mjcrockett.bsky.social keynote talk at #FAccT2025 was soo good 🔥🔥🔥 She talked about how techno-optimism is really human pessimism, how DEAD benchmarks don’t capture full human capacities and feed the hype cycle, how we need to avoid monoculture & imagine new worlds together
2387
Reposted by Anna Neumann
tante @tante.cc · 23/06/2025
It's good that bigger mainstreamer people such as John Oliver are also talking about the "AI slop" era and phenomenon. I do think he was a bit too charitable though. AI slop is not just bad visuals, it's the esthetic of modern fascism. youtu.be/TWpg1RmzAbc?...
youtu.be
AI Slop: Last Week Tonight with John Oliver (HBO)
YouTube video by LastWeekTonight
07823
Reposted by Anna Neumann
Frank Pasquale @frankpasquale.bsky.social · 23/06/2025
“One major culprit behind the wave of depersonalisation is the widespread reduction of individuals to data. Feeling invisible can stem from repeatedly experiencing standardised interactions” aeon.co/essays/our-c...
aeon.co
Our crisis is not loneliness but human beings becoming invisible | Aeon Essays
Our crisis of work and technology is one in which too many people feel that nobody sees them as a fellow human being
095
Reposted by Anna Neumann
ACM FAccT @facct.bsky.social · 24/06/2025
Our #FAccT2025 proceedings are out!! 🎉 Read all of the published papers here: dl.acm.org/doi/proceedi...
03119
Reposted by Anna Neumann
tante @tante.cc · 23/06/2025
"I do not need the one magic machine that claims to solve all my issues and then makes me jump through conversational hoops to get a mediocre result. That is actually the opposite of what I need." On chatbots as a bad design paradigm “ChatBot” is bad design
25010
Reposted by Anna Neumann
ACM FAccT @facct.bsky.social · 21/06/2025
Looking for posts about #FAccT2025? Check out our 🦋 custom feed 🦋 which is already lively and full of papers, events, and attendees for this year's conference in Athens! Click the pin 📌 in the upper right hand corner to keep this feed quickly accessible. bsky.app/profile/mari...
02812
Anna Neumann @annaneumann.bsky.social · 23/06/2025
Please come see us at the RC Trust Networking Event! You can sign up with the QR Codes around the venues and get some free drinks! 🙂‍↕️ #FAccT2025
143
Reposted by Anna Neumann
AI & Equality @aiequality.bsky.social · 16/04/2025
🚀 We’re excited to host Anna Neumann, PhD researcher (@rctrust.bsky.social) for a #PubTalk ✨ on "Seeing What's Missing: Bringing #ResponsibleAI Research Gaps into Focus Through A Landscape Analysis" ⏰ 8 May | 3PM UTC Read more and register: community.aiequalitytoolbox.com/c/march-fed1...
032
Anna Neumann @annaneumann.bsky.social · 27/05/2025
we're hiring! ✨ looking for researchers who want to make AI accountable and trustworthy 📚 PhD position (3 years, fully funded) 🔬 Postdoc position (3 years, fully funded) high independence, interdisciplinary team, real-world impact diverse backgrounds welcome: CS, law, HCI, & more interested? 👇
221
Reposted by Anna Neumann
Hypervisible @hypervisible.blacksky.app · 07/05/2025
“Massive numbers of students are going to emerge from university with degrees, and into the workforce, who are essentially illiterate…Both in the literal sense and in the sense of being historically illiterate and having no knowledge of their own culture, much less anyone else’s.”
nymag.com
Everyone Is Cheating Their Way Through College
ChatGPT has unraveled the entire academic project.
17037821382
Reposted by Anna Neumann
tante @tante.cc · 07/05/2025
"It’s not just that [AI slop's] aesthetics are useful for Trump, its entire mode of being is useful for how his administration has governed so far" The AI Slop Presidency
071
Reposted by Anna Neumann
Dr Abeba Birhane @abeba.blacksky.app · 28/03/2025
no shit www.science.org/content/arti...
science.org
AI models miss disease in Black and female patients
Analysis of chest x-rays underscores need for monitoring artificial intelligence tools for bias, experts say
5351140
Reposted by Anna Neumann
Dr Abeba Birhane @abeba.blacksky.app · 21/02/2025
they are openly advocating for the use of physiognomy in recruitment make it stop
Human capital---encompassing cognitive skills and personality traits---is critical for labor market success, yet the personality component remains difficult to measure at scale. Leveraging advances in artificial intelligence and comprehensive LinkedIn microdata, we extract the Big 5 personality traits from facial images of 96,000 MBA graduates, and demonstrate that this novel ``Photo Big 5'' predicts school rank, compensation, job seniority, industry choice, job transitions, and career advancement. Using administrative records from top-tier MBA programs, we find that the Photo Big 5 exhibits only modest correlations with cognitive measures like GPA and standardized test scores, yet offers comparable incremental predictive power for labor outcomes. Unlike traditional survey-based personality measures, the Photo Big 5 is readily accessible and potentially less susceptible to manipulation, making it suitable for wide adoption in academic research and hiring processes. However, its use in labor market screening raises ethical concerns regarding statistical discrimination and individual autonomy.
2129974
Reposted by Anna Neumann
Dave Vetter @davidrvetter.bsky.social · 21/02/2025
The rise of gen AI, backed at every level by the far-right, suggests the arrival of an _anti-information_ age. AI offers a self-reinforcing bullshit loop that, much like a Russian disinfo operation, serves to muddy the information space and create doubt about the very nature of reality.
11363134
Anna Neumann @annaneumann.bsky.social · 06/02/2025
This is horrible on so many levels. Most probably completely cemented and unaccountable by the time anyone can get Musk out of there... www.washingtonpost.com/nation/2025/...
washingtonpost.com
Elon Musk’s DOGE is feeding sensitive federal data into AI to target cuts
U.S. DOGE Service workers embedded in the Education Department have been processing sensitive agency data using artificial intelligence.
010
Anna Neumann @annaneumann.bsky.social · 04/02/2025
Funny (actually horrifying) considering Musk can't even automate cars
000
Reposted by Anna Neumann
Marietje Schaake @marietjeschaake.bsky.social · 04/02/2025
When I wrote my book The Tech Coup, quite a few people remarked that the word ‘coup’ might be an overstatement. Today it sadly seems like an understatement ↘️ bookshop.org/p/books/the-...
bookshop.org
The Tech Coup: How to Save Democracy from Silicon Valley
How to Save Democracy from Silicon Valley
15427142
Anna Neumann @annaneumann.bsky.social · 04/02/2025
"The dominance of US firms in artificial intelligence and its global market structure depends heavily on their oligopolistic control of large-scale data extraction, which translates into global control over knowledge production and consumption." www.techpolicy.press/ready-for-th...
techpolicy.press
Ready for the Broligarchy? The Tech Accountability Community Needs a Reset | TechPolicy.Press
Burcu Kilic and Renata Ávila say new strategies and tactics are necessary to accelerate visionary national industrial policies for the digital economy.
020
Reposted by Anna Neumann
Adam Rogers @jetjocko.bsky.social · 04/02/2025
See now I’m starting to think that these people want a machine to summarize things for them because when a human does it, the human uses words like “no” and “that would hurt people” and “that’s illegal.”
264092671
Reposted by Anna Neumann
Data & Society @datasociety.bsky.social · 04/02/2025
OpenAI’s ChatGPT Gov is part of a drive to automate public service & admin using ungoverned, unaccountable AI systems. It will harm vulnerable people. Writing in Tech Policy Press in December, @alicetiara.bsky.social & @kevindeliban.bsky.social saw this coming. www.techpolicy.press/ai-cant-solv...
techpolicy.press
AI Can’t Solve Government Waste – and May Hurt Vulnerable Americans | TechPolicy.Press
If you believe public institutions should actually serve the public, this is an unacceptable bait-and-switch, write Kevin De Liban and Alice Marwick.
05130
Anna Neumann @annaneumann.bsky.social · 04/02/2025
This would kill all algorithmic bias and discrimination research in the US. Right when AI systems are incorporated into government agencies. No one there to audit governing systems, this is incredibly effective and so horrifying
120
Reposted by Anna Neumann
Dr Abeba Birhane @abeba.blacksky.app · 31/01/2025
what are the projects/initiatives you know of working on developing technologies for refusal and resistance?
86530
Reposted by Anna Neumann
Deborah Blum @deborahb.bsky.social · 31/01/2025
From a journalist friend: Just spreading the word. The CDC is purging data, so people should archive their favorite CDC datasets today, namely ones around race/ethnic diversity, LGBTQ, and reproductive health. Also health data involving climate. The youth risk behavior survey has already gone down.
15347943069
Reposted by Anna Neumann
Dr Abeba Birhane @abeba.blacksky.app · 31/01/2025
repeat after me: an AI system does NOT and CANNOT seek power. it CANNOT seek, want, & desire anything more than your toaster can. people in positions of power use AI to seek, assert, consolidate, and centralise power
15644178
Anna Neumann @annaneumann.bsky.social · 11/01/2025
taking the tech out of technosolutionism, now embracing *nosolutionism*
000
Reposted by Anna Neumann
Melanie Dusseau @melaniedusseau.bsky.social · 01/12/2024
"...there is reason to worry about cognitive automation depriving us of the opportunity to build and retain those cognitive muscles. That’s the risk of intellectual and moral deskilling. It’s already happening and I think we have to resist it." -Shannon Vallor #AIethics www.vox.com/future-perfe...
vox.com
Shannon Vallor says AI does present an existential risk — but not the one you think
AI is not a “stochastic parrot,” It’s a mirror, says the philosopher.
183
Reposted by Anna Neumann
Ana Brandusescu @anabrandusescu.bsky.social · 30/11/2024
"What is needed is self-confidence on the part of democratic government to make sure that these companies, these services, are taking their proper role within a rule of law-based system, and are not overtaking it." Deep sigh and thank you @marietjeschaake.bsky.social for your words.
theguardian.com
AI expert Marietje Schaake: ‘The way we think about technology is shaped by the tech companies themselves’
The Dutch policy director and former MEP on the unprecedented reach of big tech, the need for confident governments, and why the election of Trump changes everything
0124
Reposted by Anna Neumann
Frank Pasquale @frankpasquale.bsky.social · 01/12/2024
“I’ve been skeptical about the moral machines project, because it usually ends up just trying to crowdsource moral judgments [and train AI on them] — but the whole point is that the crowd isn’t always right!…it’s a very dangerous thing to crowdsource moral judgments.” www.vox.com/future-perfe...
vox.com
Shannon Vallor says AI does present an existential risk — but not the one you think
AI is not a “stochastic parrot,” It’s a mirror, says the philosopher.
1156
Anna Neumann @annaneumann.bsky.social · 27/11/2024
If it can be scraped, it will *inevitably* used to train AI
010
Reposted by Anna Neumann
Emily M. Bender @emilymbender.bsky.social · 22/11/2024
No, ChatGPT won't help in the classroom, won't save teachers time, and doesn't represent a set of skills students need to learn. On OpenAI's latest nonsense: buttondown.com/maiht3k/arch...
buttondown.com
ChatGPT Has No Place in the Classroom
By Emily On November 20, 2024, OpenAI and an outfit called "Common Sense Media" released a guide to using ChatGPT in K-12 education—a guide which shows a...
41515191
Anna Neumann @annaneumann.bsky.social · 21/11/2024
"However, transparency does not necessarily guarantee accountability or democratic control over one’s data. While transparency is important, it is certainly not enough." www.insidehighered.com/opinion/view...
insidehighered.com
AI is consolidating corporate power in higher ed (opinion)
AI is consolidating corporate power in higher education, Lindsay Weinberg writes.
000