- nexa - server-nexa.polito.it

"A major AI training data set contains millions of examples of personal data"
by J.C. DE MARTIN July 24, 2025

July 24, 2025

*A major AI training data set contains millions of examples of personal data* /Personally identifiable information has been found in DataComp CommonPool, one of the largest open-source data sets used to train image generation models./ Eileen Guo July 18, 2025 Millions of images of passports, credit cards, birth certificates, and other documents containing personally identifiable information are likely included in one of the biggest open-source AI training sets, new research has found. Thousands of images—including identifiable faces—were found in a small subset of DataComp CommonPool, a major AI training set for image generation scraped from the web. Because the researchers audited just 0.1% of CommonPool’s data, they estimate that the real number of images containing personally identifiable information, including faces and identity documents, is in the hundreds of millions. The study that details the breach was published on arXiv earlier this month <https://arxiv.org/pdf/2506.17185>. [...] continua qui: https://www.technologyreview.com/2025/07/18/1120466/a-major-ai-training-dat…

1 0

White House Unveils America’s AI Action Plan
by Daniela Tafani July 24, 2025

July 24, 2025

White House Unveils America’s AI Action Plan The White House July 23, 2025 The White House today released “Winning the AI Race: America’s AI Action Plan”, in accordance with President Trump’s January executive order on Removing Barriers to American Leadership in AI. Winning the AI race will usher in a new golden age of human flourishing, economic competitiveness, and national security for the American people. The Plan identifies over 90 Federal policy actions across three pillars – Accelerating Innovation, Building American AI Infrastructure, and Leading in International Diplomacy and Security – that the Trump Administration will take in the coming weeks and months. Key policies in the AI Action Plan include: - Exporting American AI: The Commerce and State Departments will partner with industry to deliver secure, full-stack AI export packages – including hardware, models, software, applications, and standards – to America’s friends and allies around the world. - Promoting Rapid Buildout of Data Centers: Expediting and modernizing permits for data centers and semiconductor fabs, as well as creating new national initiatives to increase high-demand occupations like electricians and HVAC technicians. - Enabling Innovation and Adoption: Removing onerous Federal regulations that hinder AI development and deployment, and seek private sector input on rules to remove. - Upholding Free Speech in Frontier Models: Updating Federal procurement guidelines to ensure that the government only contracts with frontier large language model developers who ensure that their systems are objective and free from top-down ideological bias. “America’s AI Action Plan charts a decisive course to cement U.S. dominance in artificial intelligence. President Trump has prioritized AI as a cornerstone of American innovation, powering a new age of American leadership in science, technology, and global influence. This plan galvanizes Federal efforts to turbocharge our innovation capacity, build cutting-edge infrastructure, and lead globally, ensuring that American workers and families thrive in the AI era. We are moving with urgency to make this vision a reality,” said White House Office of Science and Technology Policy Director Michael Kratsios. “Artificial intelligence is a revolutionary technology with the potential to transform the global economy and alter the balance of power in the world. To remain the leading economic and military power, the United States must win the AI race. Recognizing this, President Trump directed us to produce this Action Plan. To win the AI race, the U.S. must lead in innovation, infrastructure, and global partnerships. At the same time, we must center American workers and avoid Orwellian uses of AI. This Action Plan provides a roadmap for doing that,” said AI and Crypto Czar David Sacks. “Winning the AI Race is non-negotiable. America must continue to be the dominant force in artificial intelligence to promote prosperity and protect our economic and national security. President Trump recognized this at the beginning of his administration and took decisive action by commissioning this AI Action Plan. These clear-cut policy goals set expectations for the Federal Government to ensure America sets the technological gold standard worldwide, and that the world continues to run on American technology,” said Secretary of State and Acting National Security Advisor Marco Rubio. <https://www.whitehouse.gov/articles/2025/07/white-house-unveils-americas-ai…>

1 0

Comment les entreprises de la tech nous forcent à utiliser l'IA
by Daniela Tafani July 23, 2025

July 23, 2025

L'IA accroît fortement les impacts environnementaux du numérique. Pour ne citer que quelques chiffres, 80% des data centers pourraient être désormais obsolètes et au cours du premier semestre 2023 l'Amérique du Nord a établi un nouveau record avec une augmentation de +25 % de construction de nouveaux centres de données. Poussées par des investissements massifs qu'il faut rentabiliser, par la promesse de nouvelles sources de profits ou par la peur de se voir dépassées, de nombreuses entreprises du numérique ont commencé à mettre en œuvre des fonctionnalités alimentées par l'IA. L'ajout de nouvelles fonctionnalités dans des applications ou des logiciels n'est bien sûr pas nouvelle en soi, mais ce qui frappe avec l'IA, c'est la manière dont les entreprises poussent à l'adoption de ces fonctionnalités, en particulier par le biais du design. Il nous semble que ce phénomène est inédit dans l'histoire des interfaces. Notre objectif avec cette enquête est d'analyser les multiples stratagèmes mis en place par les entreprises pour nous inciter à adopter l'IA, qu'on le veuille ou non. nalyse Au début de l'année 2024, nous avons commencé à remarquer que de nombreuses icônes d'IA apparaissaient dans les différents logiciels, sites web et applications que nous utilisions. Nous avons donc commencé à collecter systématiquement des captures d'écran de chaque changement d'interface lié à l'IA. Grâce au bouche à oreille et à un appel sur Mastodon, nous avons collecté plus d'une centaine de captures d'écran, tant dans les logiciels professionnels de productivité que dans d'autres types d'utilisation (loisirs, communication, créativité, etc.). En analysant ce corpus, nous constatons que les fonctionnalités basées sur l'IA sont proposées aux utilisateurs·ices de manière agressive, par diverses stratégies récurrentes. Lorsque nous comparons la manière dont les fonctions alimentées par l'IA sont introduites dans l'interface par rapport aux changements non liés à l'IA, le contraste est souvent saisissant et montre à quel point les fonctions d'IA occupent une place bien particulière. Voici donc comment les entreprises nous forcent à utiliser l'Intelligence Artificielle. Continua qui: <https://limitesnumeriques.fr/travaux-productions/ai-forcing> Qui la traduzione inglese: <https://limitesnumeriques.fr/travaux-productions/ai-forcing/en>

1 0

Re: [nexa] la questione dei commons e l'Internet, Era: "Stefano Rodotà (appositamente richiamato, dall’Aldilà mentre giocava a scacchi)"
by karlessi July 22, 2025

July 22, 2025

> Message: 1 > Date: Tue, 22 Jul 2025 10:02:54 +0200 > From: Andrea Trentini <ego(a)atrent.it> > To: nexa(a)server-nexa.polito.it > Subject: Re: [nexa] "Stefano Rodotà (appositamente richiamato > dall’Aldilà mentre giocava a scacchi)" > Message-ID: <be66ff6f-63e1-43d5-b87c-b3b49443d385(a)atrent.it> > Content-Type: text/plain; charset="utf-8" > > On 22/07/2025 09:51, Marco A. Calamari wrote: >> >> Il fatto che un componente del Garante venga attaccato in maniera pubblica mi sembra abbia una >> rilevanza generale. >> Il problema è appunto che diffondere il documento in lista amplierebbe la diffusione dello stesso, >> per cui lo vedrei inopportuno. anche io. ho ricevuto la fantomatica mail dall'indirizzo rodota.dallaldila(a)libero.it che invia evidentemente a una lista in CCN, oltre che a se stesso, all'indirizzo ricevente rodota.dallaldila(a)gmail.com ringrazio chi si è presa la briga di risparmiarmi anche solo di scorrerla (mi era bastata l'immagine fakestudioghibli a farmi passare la voglia, oltre allo screenshot della prima pagina nel corpo della mail) >> >> Decidete se, date le abitudini della lista, l'argomento merita di essere discusso in questo modo >> "asimmetrico" e non paritario. >> Io personalmente sarei favorevole. > personalmente mi interessa invece la questione commons e Internet, anzi, commons e tecnologia in generale. Tecnologie appropriate, conviviali, ecc. Tempo fa ho tradotto un libretto in merito (ai commons, non a Internet), rendendo "commons" con "beni collettivi". Ecco perché in breve: https://www.eleuthera.it/materiale.php?op=3299 la questione mi sembra rilevante per le risorse tecnologiche e via discorrendo. Perché un conto è sostenere che Internet è un "bene comune" nell'accezione della Commissione Rodotà, per cui: "In ogni caso deve essere garantita la loro fruizione collettiva, nei limiti e secondo le modalità fissati dalla legge" [1] tutt'altra questione è il "bene collettivo" nel senso a mio parere inteso da Elinor Ostrom quando analizza i "commons": - non vi è alcuna legge stabilita garante a regolare i commons - vi sono dinamiche di regolamentazione complessa, gestite da individui e comunità direttamente implicate nell'accesso, fruizione, distribuzione, manutenzione, sanzione eventuale, esclusione, ecc. - negli esempi riferiti da Ostrom per sostanziare il concetto di "commons", quindi, non c'è quindi ALCUN livello politico ulteriore rispetto all'autogestione. Non c'è un "Comune", organi dirigenziali che decidono quali ipostasi governanti il bene; tanto meno vi è un'Istituzione (Stato, Legge, ecc.) di stampo liberale e/o democratico e/o repubblicano, ecc.: ci sono le persone e le difficili relazioni fra di loro e con il loro ambiente. - i "commons" sono molto più vicini all'autogoverno che alle proposte dei "beni comuni", mi pare. Ma non sono nemmeno autogoverno, IMHO (questo aprirebbe un'altra discussione sulla necessità filosofica del governo, che esula dalla questione, almeno in parte). Sono organizzazione dal basso di ciò che interessa (inter-est: sta-tra, fra gli umani, e fra gli umani e il mondo). niente di più niente di meno. ritengo che sia cruciale, in questi tempi terribili, ragionare di commons tecnologici in questo senso. Perché il nostro ambiente ora comprende, è ovvio, anche l'ambiente tecnologico, in particolare gli ambienti tecnologici connessi a Internet. Tali "commons tecnologici" non sono quindi da intendersi come "beni comuni" nel senso della Commissione Rodotà, che interpreto come, ad esempio, "diritto all'Internet", alla connettività, ecc ecc cioè diritti fondamentali teoricamente inalienabili ma in pratica continuamente alienati, che dovrebbero essere garantiti in ultima istanza dalla legge (quella stessa legge che al momento è completamente disattesa per quanto riguarda GDPR e dintorni, fra l'altro - ma questo apre tutt'altro capitolo sullo strapotere delle multinazionali, l'insipienza criminale delle PA, il "così fan tutti", ecc.). Sono invece da intendersi come la costruzione possibile, sulla scorta degli esempi storici dell'acqua, delle risorse ittiche, forestali, ecc. ecc. studiati da Ostrom, di "beni tecnologici collettivi". Che non vuol dire collettivizzati, s'intende. Non vuol dire nemmeno gratuiti. E nemmeno liberi per tutti! L'accesso al "commons" risorse di pesca dell'area tal dei tali, raccontava Ostrom, NON è per tutti e chiunque. Analogamente, "beni tecnologici collettivi" dovrebbero essere accessibili solo a chi è intitolato a farlo dal collettivo stesso, nelle forme regolamentate da tale collettivo (commons). La legge, semmai, arriva dopo. Ma è anche superflua, IMHO, nella misura in cui le persone sanno gestire collettivamente le risorse in quanto commons. utopie? forse. ma pur sempre meglio delle distopie realizzate in cui stiamo vivendo. anche solo nel settore tech, difficile trovare delle big che non siano direttamente implicate, per dirne una, nel supporto alle guerre condotte dal governo israeliano: https://www.whoprofits.org/companies/all buona giornata k. PS: ripubblicare l'opera fondamentale del premio Nobel Ostrom in italiano sarebbe il minimo. Esaurita da parecchio, ho dovuto ricorrere a una copia della biblioteca della fondazione Basso di Roma [1] https://www.giustizia.it/giustizia/it/mg_1_12_1.page?contentId=SPS47624 art. 1 / 3c > > +1 > > p.s. ho ricevuto anche io la mail da "rodotà" >

1 0

[Responsible AI Talk] The Parity Benchmark, Shmona Simpson, CEO and Founder of Paritii
by Daniele Quercia July 22, 2025

July 22, 2025

The Parity Benchmark: A Multimodal Framework for Assessing Bias in LLMs Shmona Simpson, CEO and Founder of Paritii<https://www.paritii.com/> Join the meeting <https://teams.microsoft.com/l/meetup-join/19%3ameeting_ZTQ5NWI2MjktY2UyMy00…> Abstract: Large Language Models (LLMs) demonstrate impressive capabilities but remain susceptible to harmful biases that can reinforce inequity across domains like content moderation, healthcare, and decision-making. The Parity Benchmark offers a rigorous, multimodal framework to evaluate such biases across protected characteristics including gender, race, socioeconomic status, and intersectional identities. Using an expert-curated dataset, the benchmark assesses leading LLMs such as Granite 3.8 against GPT-4o, Claude 3.5 Sonnet, Llama 3, DeepSeekR-1, and Gemini. The data will show which LLM’s perform best across categories such as ageism, supremacism, colorism, homophobia, disability and neurodivergence. Future iterations of the benchmark will expand to include image-based evaluations, open-ended free-text prompts, and integrated multimodal analytics—providing a richer, real-world assessment of AI fairness. Initial results show that models excel differently across tasks, suggesting opportunities for targeted bias mitigation. The Parity Benchmark aims to serve as a practical, evolving tool for researchers, developers, and policymakers committed to building more equitable and accountable AI systems. <https://arxiv.org/abs/2407.02191>Bio: Dr. Shmona Simpson is the visionary CEO and Founder of Paritii<https://www.paritii.com/>, a pioneering platformdedicated to advancing equity in technology through data-driven frameworks and innovation. A recognized leader in AI ethics and technology strategy, Shmona specializes in developing solutions that promote fairness, accountability, and transparency within AI systems. Her recent development of The Parity Benchmark<https://link.springer.com/article/10.1007/s43681-024-00613-4>, published by Springer in 2025, underscores her commitment to Responsible AI. This groundbreaking framework provides a robust, data-informed approach to evaluating and enhancing fairness within machine learning systems. By identifying and mitigating biases, The Parity Benchmark equips researchers, developers, and organizations with actionable insights to create competitive, responsible AI technologies. Under Shmona’s leadership, Paritii has emerged as a trusted resource for organizations aiming to build ethical AI systems that serve broad communities. Her expertise in AI ethics and strategic innovation continues to shape critical conversations about how technology can be harnessed to promote equity and accountability. Swyft, Paritii's flagship AI solution for broad access higher education institutions, streamlines academic operations with intelligent automation. Swyft handles high-volume administrative tasks like registration, financial aid workflows, and compliance tracking—freeing up staff time and improving the student experience at scale. Swyft is available at institutions across the country<http://www.tryswyft.com/>. An influential voice in the tech industry, Shmona is a sought-after advisor and speaker on AI ethics, fairness frameworks, and the broader implications of responsible technology development. She has been invited to share her perspective at SXSW, ASU-GSV, MITRE, IBM, and Anita B.org. Before founding Paritii, she served as a Strategist at the Bill & Melinda Gates Foundation, where she strengthened health systems across diverse regions and led cross-functional initiatives to improve organizational impact. Shmona holds a DPhil from Oxford University, where she was a Rhodes Scholar. Subscribe to future talk announcements: Anyone outside Bell Labs can receive talk announcements by subscribing to the mailing list. To subscribe, send an empty email with the subject line "Subscribe RAI" to daniele.quercia(a)nokia-bell-labs.com<mailto:daniele.quercia@nokia-bell-labs.com>.

1 0

Re: [nexa] "Stefano Rodotà (appositamente richiamato dall’Aldilà mentre giocava a scacchi)"
by giancarlo＠buzzanca.net July 22, 2025

July 22, 2025

Buongiorno non ho ricevuto affatto il messaggio di cui si discute. E' possibile averne copia? Grazie in anticipo Giancarlo Buzzanca

1 0

AI slows down some experienced software developers, study finds
by Alberto Cammozzo July 22, 2025

July 22, 2025

<https://www.reuters.com/business/ai-slows-down-some-experienced-software-de…> SAN FRANCISCO, July 10 (Reuters) - Contrary to popular belief, using cutting-edge artificial intelligence tools slowed down experienced software developers when they were working in codebases familiar to them, rather than supercharging their work, a new study found. AI research nonprofit METR conducted the in-depth study, on a group of seasoned developers earlier this year while they used Cursor, a popular AI coding assistant, to help them complete tasks in open-source projects they were familiar with. The Reuters Tariff Watch newsletter is your daily guide to the latest global trade and tariff news. Sign up here. Before the study, the open-source developers believed using AI would speed them up, estimating it would decrease task completion time by 24%. Even after completing the tasks with AI, the developers believed that they had decreased task times by 20%. But the study found that using AI did the opposite: it increased task completion time by 19%. The study’s lead authors, Joel Becker and Nate Rush, said they were shocked by the results: prior to the study, Rush had written down that he expected “a 2x speed up, somewhat obviously.” The findings challenge the belief that AI always makes expensive human engineers much more productive, a factor that has attracted substantial investment into companies selling AI products to aid software development. AI is also expected to replace entry-level coding positions. Dario Amodei, CEO of Anthropic, recently told Axios that AI could wipe out half of all entry-level white collar jobs in the next one to five years. Prior literature on productivity improvements has found significant gains: one study found using AI sped up coders by 56%, another study found developers were able to complete 26% more tasks, in a given time. But the new METR study shows that those gains don’t apply to all software development scenarios. In particular, this study showed that experienced developers intimately familiar with the quirks and requirements of large, established open source codebases experienced a slowdown. Other studies often rely on software development benchmarks for AI, which sometimes misrepresent real-world tasks, the study’s authors said. The slowdown stemmed from developers needing to spend time going over and correcting what the AI models suggested. “When we watched the videos, we found that the AIs made some suggestions about their work, and the suggestions were often directionally correct, but not exactly what's needed,” Becker said. The authors cautioned that they do not expect the slowdown to apply in other scenarios, such as for junior engineers or engineers working in codebases they aren’t familiar with. Still, the majority of the study’s participants, as well as the study’s authors, continue to use Cursor today. The authors believe it is because AI makes the development experience easier, and in turn, more pleasant, akin to editing an essay instead of staring at a blank page. “Developers have goals other than completing the task as soon as possible,” Becker said. “So they’re going with this less effortful route.”

1 0

La France ne contrôle plus ses données : Microsoft les remettra aux États-Unis “si nous y sommes contraints”
by Daniela Tafani July 20, 2025

July 20, 2025

La Francia. La France ne contrôle plus ses données : Microsoft les remettra aux États-Unis “si nous y sommes contraints” Par Aymeric Geoffre-Rouland Lors d’une audition publique au Sénat, Microsoft France a confirmé qu’elle ne pouvait empêcher la justice américaine d’accéder aux données hébergées en France. Un aveu glaçant, qui révèle l’ampleur du décalage entre les promesses de souveraineté numérique et la réalité contractuelle de l’État français. Le 10 juin 2025, la commission d’enquête sénatoriale sur la commande publique a reçu Microsoft France pour une audition très attendue. Ce qui s’y est dit n’a laissé aucune place à l’ambiguïté : le droit américain s’impose, même lorsque les données sont hébergées à Paris ou Marseille. Et l’aveu n’est pas venu d’un militant ou d’un expert extérieur, mais du directeur juridique de Microsoft France lui-même. Cloud Act : Microsoft ne peut pas empêcher l’accès du gouvernement américain aux données françaises Face aux sénateurs, Anton Carniaux, directeur juridique de Microsoft France, n’a pas tourné autour du pot. À la question de savoir si Microsoft pouvait protéger les données françaises d’une injonction américaine, il a répondu, sans détour : "Si nous sommes contraints par une décision de justice américaine, nous devons remettre les données." Sénat, 10 juin 2025, vidéo à 1 h 18 m 35 s Le Cloud Act, adopté aux États-Unis en 2018, oblige toute entreprise américaine à répondre à une réquisition judiciaire, y compris pour des données stockées à l’étranger. Microsoft, bien que disposant de centres en France, reste juridiquement soumise à ce cadre. Le lieu de stockage ne fait donc pas barrière au droit.Ce qui trouble davantage, c’est que l’État français continue d’acheter massivement des services Microsoft via l’UGAP, notamment par le marché “multi-éditeurs logiciels”, dans lequel Microsoft Ireland agit en tant que fournisseur. Des milliers d’administrations, d’hôpitaux ou de collectivités utilisent ainsi Microsoft 365 ou Azure, souvent sans conscience des conséquences juridiques. Même si l’hébergement est local, le risque reste transatlantique. Et le problème ne vient pas seulement de Microsoft. Lors de la même audition, la DINUM (Direction interministérielle du numérique) a reconnu que les exigences de souveraineté définies par l’État ne sont pas encore appliquées de manière complète. Autrement dit, les marchés publics continuent de passer par des solutions non conformes, malgré la doctrine officielle. Les conditions de souveraineté, telles qu’elles ont été définies dans la doctrine de l’État, ne sont pas encore appliquées de manière complète. DINUM, Sénat, 10 juin 2025 Microsoft, de son côté, met en avant des garde-fous : contestation des demandes abusives, rapports de transparence, chiffrement avancé, projet de cloud souverain avec gouvernance européenne. Mais même ces dispositifs, aussi sérieux soient-ils, s’effacent dès lors qu’un tribunal américain prononce une injonction. Lyon : passage à l’acte pour la souveraineté numérique En réaction à ces failles, la ville de Lyon a décidé de rompre avec la suite Microsoft Office. Selon la mairie, cette décision s’inscrit “dans un contexte de prise de conscience croissante des enjeux de souveraineté numérique ”. Elle adopte la suite libre Territoire Numérique Ouvert, développée avec le SITIV, hébergée dans des datacenters régionaux, et remplaçant progressivement Microsoft par OnlyOffice, Linux et PostgreSQL.Lyon fait office de cas d’école : une collectivité publique qui met en œuvre des solutions localisées, libres et maîtrisées, afin d’échapper au poids des lois extraterritoriales. <https://www.lesnumeriques.com/societe-numerique/la-france-ne-controle-plus-…> Source : Sénat – Compte rendu de la commission d’enquête sur la commande publique numérique (10 juin 2025) <https://www.senat.fr/compte-rendu-commissions/20250609/ce_commande_publique…>

6 5

AAAI Launches AI-Powered Peer Review Assessment System
by Daniele Gambetta July 20, 2025

July 20, 2025

non so se era già passato, ma in questi giorni tra colleghi se ne parla molto.. https://aaai.org/aaai-launches-ai-powered-peer-review-assessment-system/ d

2 1

risposta a Trump per i dazi sul vino
by Angelo Raffaele Meo July 20, 2025

July 20, 2025

dazi europei sui prodotti informatici americani (Microsoft, Google, Apple, ecc. ecc.), associata al rispetto del Codice dell'Amministrazione Digitale che è norma li legge italiana assolutamente ignorata nel nostro Paese. Raf Meo

4 3