George Kaissis
LightNews LightNews
g-k.ai
George Kaissis
@g-k.ai
63 followers 71 following 14 posts
Professor for Human-Centred Transformative AI @ Hasso-Plattner-Institut. Previously @ Google DeepMind, Imperial College London, TU Munich. 🇪🇺 🏳️‍🌈

https://www.g-k.ai
Posts Replies Media Videos
Reposted by George Kaissis
ja-schnabel.bsky.social
Let’s today not only commemorate Rosalind Franklin, but also Raymond Gosling, her @kingscollegelondon.bsky.social postgraduate who took photo 51…
November 8, 2025 at 5:52 PM
g-k.ai
I am overjoyed to announce that I have joined HPI as a full professor for Human-Centred Transformative AI. I am looking forward to working with my amazing new and current collaborators and would like to deeply thank everyone who has been part of the journey that led me here! @hpi.bsky.social
November 4, 2025 at 6:04 AM
Reposted by George Kaissis
hpi.bsky.social
Zum 1. November trat Prof. Dr. med. Georg Kaissis seine Professur für #DigitalHealth: Human-Centred Transformative AI an. Sein Forschungsschwerpunkt: die Entwicklung der nächsten Generation multimodaler #KI-Modelle.
Mehr Infos zur neuen Professur: hpi.de/artikel/prof...
Neue Professur für #DigitalHealth: Prof. Dr. med. Georg Kaissis
November 3, 2025 at 10:37 AM
Reposted by George Kaissis
tum-aim-lab.bsky.social
We celebrated the 5th anniversary of our research chair at @tum.de! 💙🥂

It's been an incredible journey of research and collaboration. Thank you to everyone who has made this possible. We are very much looking forward to the next years to come!

#AIMAnniversary #AIMNews
November 3, 2025 at 12:28 PM
Reposted by George Kaissis
dppapers.bsky.social
VaultGemma: A Differentially Private Gemma Model

Amer Sinha, Thomas Mesnard, Ryan McKenna, Daogao Liu, Christopher A. Choquette-Choo, Yangsibo Huang, Da Yu, George Kaissis, Zachary Charles, Ruibo Liu, Lynn Chua, Pritish Kamath, Pasin Manurangsi, Steve He, Chiyuan ...
http://arxiv.org/abs/2510.15001
VaultGemma: A Differentially Private Gemma Model Amer Sinha, Thomas Mesnard, Ryan McKenna, Daogao Liu, Christopher A. Choquette-Choo, Yangsibo Huang, Da Yu, George Kaissis, Zachary Charles, Ruibo Liu, Lynn Chua, Pritish Kamath, Pasin Manurangsi, Steve He, Chiyuan Zhang, Badih Ghazi, Borja De Balle Pigem, Prem Eruvbetine, Tris Warkentin, Armand Joulin, Ravi KumarAmer Sinha, Thomas Mesnard, Ryan McKenna, Daogao Liu, Christopher A. Choquette-Choo, Yangsibo Huang, Da Yu, George Kaissis, Zachary Charles, Ruibo Liu, Lynn Chua, Pritish Kamath, Pasin Manurangsi, Steve He, Chiyuan Zhang, Badih Ghazi, Borja De Balle Pigem, Prem Eruvbetine, Tris Warkentin, Armand Joulin, Ravi Kumar http://arxiv.org/abs/2510.15001 We introduce VaultGemma 1B, a 1 billion parameter model within the Gemma family, fully trained with differential privacy. Pretrained on the identical data mixture used for the Gemma 2 series, VaultGemma 1B represents a significant step forward in privacy-preserving large language models. We openly release this model to the community
October 20, 2025 at 3:48 AM
g-k.ai
It has been a privilege to work with so many amazing colleagues across Google and Google DeepMind to build VaultGemma, an LLM trained from scratch with Differential Privacy. Weights are openly available! Check it out here: research.google/blog/vaultge...
VaultGemma: The world's most capable differentially private LLM
research.google
September 17, 2025 at 5:25 PM
Reposted by George Kaissis
zeynepakata.bsky.social
Due to physical resource constraints, we currently estimate that around 300–400 of the candidate papers recommended for acceptance by the ACs will need to be rejected. We seek the support of our 41 SACs in addressing this distributed optimization problem in a fair and professional manner.
August 28, 2025 at 4:12 PM
Reposted by George Kaissis
zeynepakata.bsky.social
NeurIPS has decided to do what ICLR did: As a SAC I received the message 👇 This is wrong! If the review process cannot handle so many papers, the conference needs yo split instead of arbitrarily rejecting 400 papers.
August 28, 2025 at 4:12 PM
Reposted by George Kaissis
rieck.mlsec.org
📣 Researchers in AI security, privacy & fairness: It's time to share your latest work!

The SaTML 2026 submission site is live 👉 hotcrp.satml.org

🗓️ Deadline: Sept 24, 2025

@satml.org
SaTML 2026
hotcrp.satml.org
August 27, 2025 at 6:22 PM
Reposted by George Kaissis
neuripsconf.bsky.social
NeurIPS is endorsing EurIPS, an independently-organized meeting which will offer researchers an opportunity to additionally present NeurIPS work in Europe concurrently with NeurIPS.

Read more in our blog post and on the EurIPS website:
blog.neurips.cc/2025/07/16/n...
eurips.cc
eurips.cc
A NeurIPS-endorsed conference in Europe held in Copenhagen, Denmark
eurips.cc
July 16, 2025 at 10:05 PM
Reposted by George Kaissis
bogdankulynych.bsky.social
New preprint on the most precise as of yet mapping between differential privacy and common operational notions of privacy risk used in practice:
dppapers.bsky.social Differential Privacy Papers @dppapers.bsky.social · Jul 10
Unifying Re-Identification, Attribute Inference, and Data Reconstruction Risks in Differential Privacy

Bogdan Kulynych, Juan Felipe Gomez, Georgios Kaissis, Jamie Hayes, Borja Balle, Flavio du Pin Calmon, Jean Louis Raisaro

http://arxiv.org/abs/2507.06969
Unifying Re-Identification, Attribute Inference, and Data Reconstruction Risks in Differential Privacy Bogdan Kulynych, Juan Felipe Gomez, Georgios Kaissis, Jamie Hayes, Borja Balle, Flavio du Pin Calmon, Jean Louis Raisaro http://arxiv.org/abs/2507.06969 Differentially private (DP) mechanisms are difficult to interpret and calibrate because existing methods for mapping standard privacy parameters to concrete privacy risks -- re-identification, attribute inference, and data reconstruction -- are both overly pessimistic and inconsistent. In this work, we use the hypothesis-testing interpretation of DP ($f$-DP), and determine that bounds on attack success can take the same unified form across re-identification, attribute inference, and data reconstruction risks. Our unified bounds are (1) consistent across a multitude of attack settings, and (2) tunable, enabling practitioners to evaluate risk with respect to arbitrary (including worst-case) levels of baseline risk. Empirically, our results are tighter than prior methods using $\varepsilon$-DP, R\'enyi DP, and concentrated DP. As a result, calibrating noise using our bounds can reduce the required noise by 20% at the same risk level, which yields, e.g., more than 15pp accuracy increase in a text classification task. Overall, this unifying perspective provides a principled framework for interpreting and calibrating the degree of protection in DP against specific levels of re-identification, attribute inference, or data reconstruction risk.
July 10, 2025 at 10:05 AM
Reposted by George Kaissis
dppapers.bsky.social
The Hitchhiker's Guide to Efficient, End-to-End, and Tight DP Auditing

Meenatchi Sundaram Muthu Selva Annamalai, Borja Balle, Jamie Hayes, Georgios Kaissis, Emiliano De Cristofaro

http://arxiv.org/abs/2506.16666
The Hitchhiker's Guide to Efficient, End-to-End, and Tight DP Auditing Meenatchi Sundaram Muthu Selva Annamalai, Borja Balle, Jamie Hayes, Georgios Kaissis, Emiliano De Cristofaro http://arxiv.org/abs/2506.16666 This paper systematizes research on auditing Differential Privacy (DP) techniques, aiming to identify key insights into the current state of the art and open challenges. First, we introduce a comprehensive framework for reviewing work in the field and establish three cross-contextual desiderata that DP audits should target--namely, efficiency, end-to-end-ness, and tightness. Then, we systematize the modes of operation of state-of-the-art DP auditing techniques, including threat models, attacks, and evaluation functions. This allows us to highlight key details overlooked by prior work, analyze the limiting factors to achieving the three desiderata, and identify open research problems. Overall, our work provides a reusable and systematic methodology geared to assess progress in the field and identify friction points and future directions for our community to focus on.
June 23, 2025 at 3:48 AM
Reposted by George Kaissis
tum-aim-lab.bsky.social
Responsible medical AI demands patient-level privacy. Our recent #TPDP '25 paper extends Differential Privacy #DP beyond individual data points to protect entire patient profiles. 🧠

📄 Read the paper: tinyurl.com/k9fz456a
🎥 Watch the video: tinyurl.com/2jx528m5
tpdp.journalprivacyconfidentiality.org
June 12, 2025 at 7:55 AM
Reposted by George Kaissis
paulhager.bsky.social
Going to be at CVPR the next couple of days presenting our paper „A Tale of Two Classes: Adapting Supervised Contrastive Learning to Binary Imbalanced Datasets“.

arxiv.org/abs/2503.17024

Always happy to meet anyone working on representation learning or tabular DL and medical data
A Tale of Two Classes: Adapting Supervised Contrastive Learning to Binary Imbalanced Datasets
Supervised contrastive learning (SupCon) has proven to be a powerful alternative to the standard cross-entropy loss for classification of multi-class balanced datasets. However, it struggles to learn ...
arxiv.org
June 11, 2025 at 12:56 AM
Reposted by George Kaissis
hpi.bsky.social
Markiert es euch im Kalender: Die IOI 2027 kommt ans HPI! 🎉

Zum ersten Mal seit 1992 kommt die Internationale Informatik-Olympiade (IOI) zurück nach Deutschland und wir sind in Kooperation mit Bundesweite Informatikwettbewerbe (BWINF) stolze Gastgeber des renommierten Wettbewerbs.
Grafik die auf die Internationale Informatik Olympiade 2027 hinweist. Datum: 12.–19. September 2027 Grafik die auf die Internationale Informatik Olympiade 2027 hinweist. Datum: 12.–19. September 2027 Grafik die auf die Internationale Informatik Olympiade 2027 hinweist. Datum: 12.–19. September 2027
May 30, 2025 at 10:32 AM
g-k.ai
Check out our new pre-print "Strong Membership Inference Attacks on Massive Datasets and (Moderately) Large Language Models", joint work with fantastic colleagues from Google (DeepMind) and many other great institutions! Find it here: arxiv.org/abs/2505.18773
Strong Membership Inference Attacks on Massive Datasets and (Moderately) Large Language Models
State-of-the-art membership inference attacks (MIAs) typically require training many reference models, making it difficult to scale these attacks to large pre-trained language models (LLMs). As a resu...
arxiv.org
May 27, 2025 at 8:00 AM
g-k.ai
Check out our new pre-print "Redirection for Erasing Memory (REM): Towards a universal unlearning method for corrupted data", joint work with excellent colleagues from Google DeepMind, Google Research and University of Cambridge. Check it out here! arxiv.org/abs/2505.17730
Redirection for Erasing Memory (REM): Towards a universal unlearning method for corrupted data
Machine unlearning is studied for a multitude of tasks, but specialization of unlearning methods to particular tasks has made their systematic comparison challenging. To address this issue, we propose...
arxiv.org
May 26, 2025 at 7:17 AM
Reposted by George Kaissis
tum-aim-lab.bsky.social
We are incredibly proud of Prof @danielrueckert.bsky.social for being elected as Fellow of the Royal Society! Well deserved and a testament to your dedication to research 🎉
royalsociety.org The Royal Society @royalsociety.org · May 20
We are pleased to announce the 90 outstanding researchers from across the world who have been elected to the Fellowship of the Royal Society this year. The group includes trailblazers from AI and electron microscopy to global health and neuroscience. #RSFellows royalsociety.org/news/2025/05...
Exceptional scientists elected as Fellows of the Royal Society | Royal Society
Over 90 outstanding researchers from across the world have this year been elected to the Fellowship of the Royal Society, the UK’s national academy of sciences.
royalsociety.org
May 22, 2025 at 11:08 AM
Reposted by George Kaissis
ec.europa.eu
Now, it's time to up the ante.

We are committed to enshrining scientific freedom in EU law, creating a 7-year ‘super grant’ to attract top researchers, and expanding support for the most promising scientists.

More → europa.eu/!TTbWbJ
May 14, 2025 at 7:02 AM
Reposted by George Kaissis
tum-aim-lab.bsky.social
Data attribution is crucial for debugging models and detecting low quality data (spotting mislabeled samples, bias etc.).
But many methods aren't mathematically sound and don’t scale.

But how could we improve this for large models?
1/n
April 20, 2025 at 8:49 AM
Reposted by George Kaissis
paulhager.bsky.social
Excited to share: “A Tale of Two Classes: Adapting Supervised Contrastive Learning to Binary Imbalanced Datasets” has been accepted to #CVPR2025! 🎉

Paper: lnkd.in/esKRqF5p
Code: lnkd.in/eZFvDA5Q

(Thread incoming 👇)
April 1, 2025 at 1:37 PM
Reposted by George Kaissis
serge.belongie.com
Would you present your next NeurIPS paper in Europe instead of traveling to San Diego (US) if this was an option? Søren Hauberg (DTU) and I would love to hear the answer through this poll: (1/6)
NeurIPS participation in Europe
We seek to understand if there is interest in being able to attend NeurIPS in Europe, i.e. without travelling to San Diego, US. In the following, assume that it is possible to present accepted papers ...
docs.google.com
March 30, 2025 at 6:04 PM
g-k.ai
Congratulations to my revered mentor and dear friend @danielrueckert.bsky.social for this great honour and for the outstanding and enduring achievements that underlie it!
dfg.de Deutsche Forschungsgemeinschaft @dfg.de · Mar 19
🥁 Der Moment ist gekommen! 🏅 Heute ehren wir 10 herausragende Wissenschaftler*innen mit dem #LeibnizPreis 2025 der!👏 Für ihre bahnbrechenden Arbeiten und Verdienste in ihren Fachrichtungen erhalten sie je 2,5 Mio. € zur Förderung ihrer #Forschung. Herzlichen Glückwunsch an alle Preisträger*innen!
Gruppenbild der Leibnizpreisträger*innen mit DFG-Präsidentin Katja Becker, DFG-Generalsekretärin Heide Ahrens, Wissenschaftsminister Cem Özdemir, Staatssekretär Markus Blume und Senatorin Ina Czyborra.
March 20, 2025 at 5:44 AM
Reposted by George Kaissis
dppapers.bsky.social
$( varepsilon, δ)$ Considered Harmful: Best Practices for Reporting Differential Privacy Guarantees
Juan Felipe Gomez, Bogdan Kulynych, Georgios Kaissis, Jamie Hayes, Borja Balle, Antti Honkela
http://arxiv.org/abs/2503.10945
$( varepsilon, δ)$ Considered Harmful: Best Practices for Reporting Differential Privacy Guarantees Juan Felipe Gomez, Bogdan Kulynych, Georgios Kaissis, Jamie Hayes, Borja Balle, Antti Honkela http://arxiv.org/abs/2503.10945 Current practices for reporting the level of differential privacy (DP) guarantees for machine learning (ML) algorithms provide an incomplete and potentially misleading picture of the guarantees and make it difficult to compare privacy levels across different settings. We argue for using Gaussian differential privacy (GDP) as the primary means of communicating DP guarantees in ML, with the full privacy profile as a secondary option in case GDP is too inaccurate. Unlike other widely used alternatives, GDP has only one parameter, which ensures easy comparability of guarantees, and it can accurately capture the full privacy profile of many important ML applications. To support our claims, we investigate the privacy profiles of state-of-the-art DP large-scale image classification, and the TopDown algorithm for the U.S. Decennial Census, observing that GDP fits the profiles remarkably well in all three cases. Although GDP is ideal for reporting the final guarantees, other formalisms (e.g., privacy loss random variables) are needed for accurate privacy accounting. We show that such intermediate representations can be efficiently converted to GDP with minimal loss in tightness.
March 17, 2025 at 3:33 AM
Reposted by George Kaissis
helmholtzmunich.bsky.social
#FUTURE-AI: International Experts Define Guidelines for Trustworthy Healthcare AI 📝

👉Learn more:
t1p.de/2rgfh

👉Check out the interview with Prof. Julia Schnabel & Dr. Georgios Kaissis:
t1p.de/e9g6a

#TrustworthyAI
International Experts Establish FUTURE-AI Guidelines for Trustworthy Healthcare AI
March 12, 2025 at 2:05 PM