AI Accountability Lab

we at @aial.ie, are investigating amplification/censorship on X/twitter in EU. what are major EU:
-politicians & regulators
-journalists
-influencers & public intellectuals, with high visibility/influence on X, across both left & right political ideologies

we're interested in existing datasets too

Amazing group of AI auditing scholars and practitioners at the AI Accountability Lab @aial.ie, led by @abeba.bsky.social ✨

Last but not least is Monique Munarini (she/her), a Qualified lawyer & policy analyst doing PhD at the University of Pisa- Italy who is a visiting researcher at the AIAL exploring approaches to AI audits rooted in feminist theory, participatory action research, & design justice principles.

8/

Giovanni Occhipinti, an MSc student in Language Technology at the University of Bologna and UCD, is an intern at our lab. He is developing Narratives of AGI Corpus (NAGICOR), a curated collection of publications focused on AGI aiming to enable empirical investigations of the AGI debate.

7/

Mauricio Dal Pozzo Schneider is a legal professional currently enrolled in the European Master in Law, Data and AI (EMILDAI) program, and working on European legal frameworks, specifically GDPR, AI Act, & DSA, envisioning how AI systems should be assessed regarding impacts on fundamental rights.

6/

Nana Nwachukwu (she/her), @mtechlaw.bsky.social‬ is a lawyer with a background in digital media & Int’l Affairs + over a decade of experience working at the intersection of technology, human rights, & policy. Nana's PhD focuses on AI ecosystem accountability, algorithmic governance, & audits.

5/

Riccardo Angius (they/them) @storitu.org are pursuing a PhD at the AIAL. Their past work on DSA-driven audits – featured in international conferences and newspapers – secured significant accountability outcomes. They currently work on platform auditing and corporate capture of AI.

4/

Maribeth Rauh (she/her) @mbrauh.bsky.social is a software engineer & researcher with almost a decade of experience at Google DeepMind & expertise in evaluation of AI systems. Maribeth’s current work focuses on audit frameworks and evaluations of AI companionship applications and "agents."

3/

Dr Harshvardhan Pandit @harshp.com is a computer scientist by training that works with laws like GDPR & AI Act to ensure tech is accountable. Harsh comes with experience in privacy & AI standardisation and currently works on improving impact assessments and addressing regulatory capture.

2/

We are thrilled to finally introduce our lab members aial.ie/people/

1/

We call on the Commission to refrain from pursuing a deregulation agenda and champion the proper enforcement and implementation of the AI Act and the wider EU #digital rulebook.

We firmly oppose any attempt to delay or re-open the #AIAct, particularly in light of the growing trend of deregulation, which risks undermining key accountability mechanisms & hard-won rights, enshrined in EU law, protecting people, the planet, justice and #democracy.

Open Joint Letter against the Delaying and Reopening of the AI Act CDT Europe, alongside the European Consumer Organisation (BEUC), European Digital Rights (EDRi) and the European Centre for Not-for-Profit Law (ECNL), co-drafted a letter signed by 52 civil society or...

⏰ Today, CDT Europe, together with 51 experts, academics & civil society organisations, sent an open letter to the European Commission to express our concerns regarding the forthcoming Digital Simplification package, which could include revisiting the AI Act.

👇🏻 Read the full letter on our website:

Shout out to the dream team that worked meticulously and patiently through the long journey that this paper has been over the past few years: Ria Kalluri, @willie-agnew.bsky.social, Myra Cheng, Kentrell Owens, @soldaini.net & yours truly

Following qualitative inspection of 100 CV papers & 100 patents, we highlight two salient themes that emerged:

1. humans are subsumed under the term ‘objects’
2. the field goes out of its way to obfuscate surveillance

8/

Are only a few rogue entities and authors powering to surveillance? Nop! We found, across institutions, nations and subfields, the practice of research to surveillance pipeline is a pervasive norm.

7/

a, Institutions and nations that have produced the most CVPR papers with downstream surveillance-enabling patents. For each institution or nation, most patented papers have been used in surveillance-enabling patents. b, Percentage of patented papers used in surveillance-enabling patents, stratified by institution, nation and subfield. For each institution, nation or subfield that has published at least ten papers with downstream patents, we show the percentage of these papers that have been used in surveillance-enabling patents (vertical grey bars) (n = 13,804, n = 18,272 and n = 19,413, respectively). We found a pervasive norm: if an institution, nation or subfield authors papers with downstream patents, most are used in surveillance-enabling patents. (Vertical grey bars are frequently above the 50% threshold; orange line). Whiskers represent the standard deviation.

We also looked at institutions & nations that are producing the most surveillance.
Top inst:
1.MSFT
2.CMU
3.MIT
4.Chinese Univ of Hong Kong

Top nations:
1.USA
2.China
3.UK

If an institution, nation or subfield authors papers with downstream patents, most are used in surveillance patents.

6/

☝🏾 finding indicates that not only has there been a major change towards surveillance but it is one of the most salient changes that occurred in the field over the past decades. The language & patenting practices in CV have evolved to increasingly focus on analysing humans & enabling surveillance

5/

To get insight into linguistic evolution of the field,we compared the log odds ratios of word frequencies in paper titles from 1990s vs 2010s.We see a clear qualitative shift from generic application-ambiguous language in 1990s (‘shape’, ‘edge’) to a focus in 2010s on humans (‘action’, ‘person’)

4/

a, Across three decades of patented computer-vision papers (n = 11,917), there has been a steady increase in the proportion used in surveillance-enabling patents. Whiskers represent the standard deviation. b, Number of computer-vision papers used in surveillance-enabling patents. Comparing the 1990s to the 2010s, the number of computer-vision papers used in only non-surveillant patents has remained relatively stable, whereas the number of computer-vision papers used in surveillance-enabling patents has risen more than fivefold. Whiskers represent the standard deviation. c, Differences in word frequencies between paper titles from the 1990s versus those from the 2010s. We report highly polarized words (with z scores computed using weighted log odds ratios). All word associations shown are statistically significant (P < 0.01). There is a clear qualitative shift from the more generic paper focus of the 1990s (turquoise bars) to an increased focus on analysing semantic categories and humans (for example, ‘semantic’, ‘action’ or ‘person’) in the 2010s (pink bars).

Across 3 decades, while the number of papers used in only non-surveillant patents remained relatively stable, from the 1990s to the 2010s, there was a *more than fivefold increase* in the number of computer-vision papers used in surveillance patents

3/

a, Relative frequencies of data types extracted from computer-vision papers and patents. For each year from 2010 to 2019, we randomly sampled and analysed ten paper–patent pairs (n = 200). Most of the annotated computer-vision papers and patents (88%, s.d. = 5.7%) refer to data about humans. Most of the papers and patents (68%, s.d. = 4.7%) specifically refer to data about human bodies and body parts. Only 1% (s.d. = 0.7%) of the papers and patents targeted exclusively non-socially salient data. b, Examples of images analysed in computer-vision papers. For a random sample of the computer-vision papers (n = 50), we display one example of an image analysed by the paper. For papers that analysed any images containing humans, we display an example of these images of humans (highlighted in red). For papers that did not analyse any images of humans, we display an example of these non-human-depicting images (highlighted in grey). Many papers analysed images of humans. Images in b are adapted from the following references and are, unless otherwise stated, from IEEE, under a Creative Commons licence CC BY ND. Top row, left to right. Second row, left to right. Third row, left to right. Fourth row, left to right (except second image). Fifth row, left to right. Fourth row (second image).

What type of surveillance, you ask? 37% focused on human bodies (targeting humans in the midst of everyday activities such as walking, shopping), 31% on human body parts (faces, eyes), 17% human spaces (homes, offices, roads), while only 1% of papers & 1% of patents extracted only non-human data

2/

Computer-vision research powers surveillance technology - Nature An analysis of research&nbsp;papers and citing patents indicates the extensive ties between computer-vision research and surveillance.

New paper hot off the press www.nature.com/articles/s41...

We analysed over 40,000 computer vision papers from CVPR (the longest standing CV conf) & associated patents tracing pathways from research to application. We found that 90% of papers & 86% of downstream patents power surveillance

1/

Builder.ai® - Composable Software Development Platform Get your software applications developed easily. AI makes developing your software applications on our award-winning platform faster and more cost-effective. See how it works.

Builder.ai promised to revolutionize software development with AI, backed by Microsoft and valued at $1.5 billion masked manual labor as machine learning until the facade crumbled, leaving behind lawsuits, layoffs, & one of the industry's most embarrassing collapses www.techspot.com/news/108173-...

Keeping the Web Up Under the Weight of AI Crawlers If you run a site on the open web, chances are you've noticed a big increase in traffic over the past few months, whether or not your site has been getting more viewers, and you're not alone. Operator...

AI crawlers generate so much traffic that some web sites are struggling to stay online. What can be done?

INDL_MEA_2 - International Network on Digital Labor International Network on Digital Labor is launching a dedicated chapter for the region. This conference provides a unique platform to present research related to the MEA region, both ongoing and/or bu...

📢 Call for Papers! The 2nd INDL-MEA Conference is coming November 25–26, 2025 — fully online! Join the International Network on Digital Labor to explore #PlatformWork, #AI & #GigEconomy in the Middle East & Africa. Abstracts due August 31, 2025. 🌍
👉 www.indl.network/indl_mea_2/

wonderful development

Join us tomorrow! bsky.app/profile/data...

I'm looking for a postdoc to work on justice-oriented audit frameworks. Closing date April 30, 2025. pls share widely

Research project/Challenge The AIAL (AI Accountability Lab) is seeking to appoint a Postdoctoral Researcher to work on developing a justice-oriented audit framework synthesising computational methods, theories of justice, and existing regulations to premeditatively focus audits towards meaningful accountability. The goal of the framework is to provide audit practitioners with practical tools such as guiding questions and rubrics that shape perspectives towards rigorous justice oriented audits. The position corresponds to work in one or more of the following areas: Accountability Mapping accountability mechanisms and governance structures and their alignments with fundamental rights and freedoms and legal frameworks. Challenging existing accountability mechanisms that do not consider or sufficiently address social inequalities, power and resource asymmetries. Developing new methods for ensuring accountability beyond technical and organisational considerations that provide empirical evidence for holding stakeholders accountable for AI development, provision, and deployments. Auditing Developing audit methodologies for specific stages in the AI lifecycle focused on ensuring justice, accountability, and transparency beyond merely satisfying legal requirements. Development of verifiable, replicable, and reproducible design methodologies and frameworks and using these in the execution of audits. Developing audit tools and frameworks to evaluate AI development and deployments with a specific focus on risk and harm mitigations beyond technical and organisational issues.

We are seeking to appoint a Postdoctoral Researcher to work on developing a justice-oriented audit framework synthesising computational methods, theories of justice, and existing regulations to premeditatively focus audits towards meaningful accountability. www.adaptcentre.ie/careers/post...

How advanced a society is is not measured by whether it is engaged in space travel or developes artificial intelligence.