Augustin Godinot

Algorithm Auditing | CS PhD student @ INRIA/IRISA/PEReN | Visiting UofT @ CleverHans Lab !

[Augustin Godinot @augodinot]

Interested in how AI providers can alter their products to evade regulation, the possible defenses and what it means for AI governance?
Or just want a nice soothing voice to nap after lunch?

I'm defending on Feb. 10 at 2:30PM, Paris time.
3y of PhD distilled in 45min! 🔗in🧵

[Mohammad Yaghini @MYaghini]

Can ML models be trained under fairness constraints with formal DP guarantees, without sacrificing utility?
New paper at ICLR 2026: RaCO-DP
Poster: Pavilion 4 (later today!)
🧵
(w/ @tudorcebere, Michael Menart, Aurélien Bellet, @NicolasPapernot )

[El Mahdi El Mhamdi @L_badikho]

Wassim's most important PhD contribution was to discover 𝐈𝐧𝐝𝐢𝐫𝐞𝐜𝐭 𝐃𝐚𝐭𝐚 𝐏𝐨𝐢𝐬𝐨𝐧𝐢𝐧𝐠 (poisoning with data that is absent from the training set).

As often in AI, the prior work of a PhD student gets invisibilised by announcements from large teams.

Let's fix that

[Timothée Chauvin @timotheechauvin]

New research: cheaply detecting changes in LLM APIs.

We published two papers on the topic:
- Log Probability Tracking of LLM APIs (ICLR 2026)
- Token-Efficient Change Detection in LLM APIs

Both papers request a single token of output from APIs, enabling unprecedently cheap monitoring.

[Augustin Godinot @augodinot]

@pmangold_ Merci Paul !

[Augustin Godinot @augodinot]

PhDone! A few weeks ago, I defended my contributions to avoid a dieselgate moment for AI regulation.

After some rest, I am now entering a blissful period of pure academic freedom, courtesy of the French unemployment benefits 🇫🇷 Time to look for a new academic home !

[Augustin Godinot @augodinot]

@chhaviyadav_ Thanks a lot Chhavi!

[Augustin Godinot @augodinot]

Slides and manuscript here: https://t.co/fjR05gzP3n

[Augustin Godinot @augodinot]

The defense, and all the work that was distilled into it, would not have been so enjoyable without the guidance from my advisors Gilles, Erwan, Francois, Camilla and Gohar.

[Augustin Godinot @augodinot]

To tune in remotely, see https://t.co/fjR05gzP3n
I'm also on the job market, here are some interests:
trustworthy ML, AI governance, outsider system scrutiny, platforms attempts at evading regulations

[Augustin Godinot @augodinot]

Interested in how AI providers can alter their products to evade regulation, the possible defenses and what it means for AI governance?
Or just want a nice soothing voice to nap after lunch?

I'm defending on Feb. 10 at 2:30PM, Paris time.
3y of PhD distilled in 45min! 🔗in🧵

[Sarah Cen @cen_sarah]

In the AI ecosystem, who supplies the data? the compute? the models?

We just released a new tool on the AI Supply Chain. Our dataset reveals how AI models, data, compute, capital, and even talent change hands.

Here’s why you should care 👇

[Augustin Godinot @augodinot]

We are presenting the paper with Milos tomorrow at 11am, come chat at our poster !
11am, East Exhibition Hall A-B #E-1911

[Augustin Godinot @augodinot]

📯 ICML25 spotlight 📯 How to detect and prevent audit manipulations?

Do you remember 🚗 Dieselgate 💨? The car computer would detect when it was on a test-bench and reduce the engine power to fake environmental compliance. Well, this can happen in AI too. 🧵1/6

[Augustin Godinot @augodinot]

@chhaviyadav_ See you in Vancouver!

[Augustin Godinot @augodinot]

@trustworthy_ml @icmlconf Current incentives for "Trustwortyness" are reputational. Thus, models developers only need to fake it. We present a research direction to prevent this:
https://t.co/XQzpSIoaNW

[Augustin Godinot @augodinot]

📯 ICML25 spotlight 📯 How to detect and prevent audit manipulations?

Do you remember 🚗 Dieselgate 💨? The car computer would detect when it was on a test-bench and reduce the engine power to fake environmental compliance. Well, this can happen in AI too. 🧵1/6

[Konrad Rieck 🌈 @mlsec]

We're excited to announce the Call for Papers for SaTML 2026, the premier conference on secure and trustworthy machine learning @satml_conf

We seek papers on secure, private, and fair learning algorithms and systems.

👉 https://t.co/cPFitlsXu2
⏰ Deadline: Sept 24

[Stephan Rabanser @steverab]

📢 New ICML 2025 paper!

Confidential Guardian: Cryptographically Prohibiting the Abuse of Model Abstention

🤔 Think model uncertainty can be trusted?
We show that it can be misused—and how to stop it!
Meet Mirage (our attack💥) & Confidential Guardian (our defense🛡️).

🧵1/10

[Augustin Godinot @augodinot]

- https://t.co/u3OdWWxxyl
- https://t.co/ZYsgVTY2Mb

[Augustin Godinot @augodinot]

📯 ICML25 spotlight 📯 How to detect and prevent audit manipulations?

Do you remember 🚗 Dieselgate 💨? The car computer would detect when it was on a test-bench and reduce the engine power to fake environmental compliance. Well, this can happen in AI too. 🧵1/6

[Augustin Godinot @augodinot]

🧵6/6 If you want to read more about this, I encourage you to read the paper, but not only!
Recently, there has been a lot of exciting works on robust audits, here are a few I enjoyed:
- https://t.co/ST0bum51KG
- https://t.co/hD5oslLr0t
- https://t.co/ww5C6JVnff