Nils Lukas portrait

Contact

Academic

nils.lukas@mbzuai.ac.ae

🏫 MBZUAI, Abu Dhabi, UAE

Startup

nlukas@adherence.ae

🚀 Adherence, Abu Dhabi, UAE

Company inquiries: Adherence address only.

Career

2024–nowAssistant Professor
MBZUAI
CSAC (co-director)
2019–24PhD
University of Waterloo
2024Research intern
Borealis AI
2022Research intern
Microsoft Research
2016–18MSc
RWTH Aachen

Guten Tag, I’m Nils 👋. I’m an Assistant Professor of Machine Learning at MBZUAI in Abu Dhabi, where I lead the Secure, Private, Open and Trustworthy SPOT Lab logo AI lab and co-direct the Center for Secure Agentic Control (CSAC). My group works on watermarks, adaptive attacks, and privacy and security for AI agents.

I did my PhD at the University of Waterloo, where my thesis received the Top Mathematics Doctoral Prize and I received the 2024 Alumni Gold Medal. Since 2025, my research has been awarded nearly $2 million in research grants, including an Amazon Research Award and an AED 1 million award from the MBZUAI Innovation Program. Recently I founded Adherence, a startup delivering enterprise AI agents. Check out our website.

We're hiring

Email me
Postdocs
AI Safety, Multi-Agent Systems, Self-Improvement, Differential Privacy
PhD students
Security for AI Agents, Scalable Oversight, Privacy
Teaching
Scholar GitHub LinkedIn CV

Latest News (see all)

Nov'26 We hosted the Agentic AI Security Symposium at MBZUAI, which I co-chaired. My talk was Your Agent Cannot Keep a Secret: Leakage, Side Channels and Provable Mitigations.
Oct'26 🥇 I received an award of AED 1 million from the MBZUAI Innovation Program for Prompt Internalization for Enterprise AI Agents.
Oct'26 🥇 Etihad Airways extended our conversational booking agents project to February 2027 with a further $225,000, adding red-teaming and research on open-weight models.
Oct'26 🔥 CopyShield: A Cross-Level Benchmark of Copyright Defenses in LLMs is an 🏆 Oral Presentation at EMNLP'26.
Sep'26 🔥 Two papers accepted at NeurIPS'26: Entropy-Gated Latent Recursion and PROBE: Learning to Audit Policy Compliance in Tool-Using LLM Agents.
Aug'26 🔥 Two papers accepted at EMNLP'26: CopyShield, a cross-level benchmark of copyright defenses in LLMs (main conference), and TRACER: Trace-Regularized Training Against Indirect Prompt Injection (Findings).
Jul'26 🥇 I am a PI on a grant from the United Al-Saqer group on MedBoard: Privacy-preserving Multi-Agent AI for Healthcare, with Praneeth Vepakomma and Preslav Nakov.
Jun'26 🔥 SPQR, a benchmark for the safety alignment of text-to-image models under benign fine-tuning, was accepted at ECCV'26 (Datasets & Benchmarks).
May'26 I gave a talk on 'AI Safety: Alignment, Oversight and Deception' at the ADIA Lab & UGR Summer School in Granada, Spain.
Apr'26 🔥 Two papers accepted at ICML'26: Collaborative Threshold Watermarking and CoRe: Collaborative Reasoning via Cross Teaching.
Apr'26 🔥 Two papers accepted at ACL'26, one at the main conference and one at Findings.
Feb'26 🔥 One accepted paper at CVPR'26 as an 🏆 Oral (top 0.8%).
Feb'26 I attended our US AI Forum in San Francisco, then traveled to Washington DC for a CRA career event.
Feb'26 I am a PI on a grant from TII (Technology Innovation Institute), together with Salem Lahlou, to research GFlowNets for fuzzing of agentic applications.
Jan'26 🔥 Two accepted papers. DP-Fusion to ICLR'26 and SD-E^2 to EACL (Findings).
Dec'25 🥇 I received an Amazon Research Award
Sept'25 I was invited by Hady Elsahar to give a remote talk to the watermarking team at Meta (FAIR) in Paris on 'Adaptively Robust and Forgery-Resistant Watermarking'.
Sept'25 I am a co-PI on a grant from the United Al-Saqer group to research privacy-preserving brain computer interfaces, with Abdulrahman Mahmoud and Elizabeth Churchill.
Sept'25 🔥 One paper on mask image watermarking accepted at NeurIPS’25.
Aug'25 I am a PI on a grant with Etihad Airways to research and develop conversational booking agents.
Aug'25 🔥 One paper on patching Speech Language Models accepted at EMNLP’25 (main).
May'25 🔥 Two papers accepted at ICML’25, one as a 🌟 Spotlight (top 2.6%).
May'25 I travelled to the US AI Forum by MBZUAI in San Francisco.
Dec'24 🥇First place in the NeurIPS'24 watermark removal challenge with 4.4k USD prize money. Congrats to Fahad Shamshad and the team. (link)
Oct'24 🥇First place in GITEX Cybersecurity hackathon with 10.8k USD prize money. Congrats to my MSc student Tameem Bakr and the team.
Aug'24 I joined MBZUAI as an Assistant Professor of Machine Learning.
Jun'24 🔥 Two papers accepted at USENIX’24 on Private Set Intersection & secure DNN inference.
Apr'24 🥇Received the Mathematics Doctoral Prize and the Waterloo Alumni Gold Medal.
Jan'24 🔥 Two papers accepted at ICLR’24 (Vienna) on adaptive image‑watermark attacks & universal backdoors.
Dec'23 Gave a research talk at Meta on privacy of personal information in fine‑tuned LLMs.
Oct'23 Talk at UC Berkeley & 🥇Best Poster Award at Cheriton Symposium (Waterloo).
Oct'23 🏆 Distinguished Contribution Award at Microsoft MLADS for our IEEE S&P’23 paper.
Jun'23 Research talks at Google & MongoDB; IEEE S&P’23 paper featured on the Microsoft Research Focus blog.
May'23 🔥 Paper on learnable watermarks for image generators accepted at USENIX’23 Security (Anaheim).
Apr'23 🔥 Paper on PII leakage in language models accepted at IEEE S&P’23 (San Francisco).
May'22 Started research internship at Microsoft Research with Shruti Tople & Lukas Wutschitz.
Feb'22 🏆 Awarded the David R. Cheriton Scholarship for outstanding academic achievements.

Research map 29 papers in 4 topics, 2020 to 2026