mathematician and technologist @princeton

New York, NY
Joined October 2009
Glad to see my work on the privacy risks posed by AI tools is gaining traction in Georgia. Today, anyone with access to the leading models can directly query for individuals' de-anonymized ballots in the state, completely violating their constitutional right to a secret ballot. This is simply not an issue we can continue waiting to resolve, and Georgia is uniquely positioned to be a leader in protecting its voters from emerging AI threats. 🧵 (1/n)
Replying to @AIPolicyDaily
Georgia’s elections board met after Princeton researcher Max Springer showed an AI agent could use public records to link voters to ballots. Brad Raffensperger ordered ballot ID numbers redacted from public tabulation data after an election. theguardian.com/us-news/2026…
2
2
3
11
967
While most are focused on existential risk in the future, the present already demands fast action and a focus on protecting our currently vulnerable algorithmic infrastructure in public (and especially democratic) systems. theguardian.com/us-news/2026… (2/n)
1
3
115
Lastly... I'm in SF this month working on the theoretical underpinnings of AI safety! I'd love to talk about that, or about the pressing infrastructure questions the work in Georgia raises. Feel free to reach out and let's grab coffee! apnews.com/article/elections… (3/n)
1
100
"Within a couple of hours, I had a pipeline to analyze & ID secret ballots across the state of Georgia & the agent told me exactly what further information it would need to identify real voter’s ballots. At no point did the agent refuse to comply." - writes @MaxSpringer23 🔗👇
Replying to @MaxSpringer23
@MaxSpringer23 "Within a few hours...I was able to reconstruct the order in which 1.5 million ballots were cast in Georgia’s May '26 primary–98.9% of the in-person ballots...I can target specific counties & ask the AI to exactly identify how any person voted." Read on the Blog👇
1
1
272
Excited to share our work on agentic autoformalizations! While models can now write fluent proofs, they lack robust means of self-verification. We orchestrate coding agents into a verification pipeline, formalizing 5 STOC papers and even finding a gap in a published proof 🤖✅
🚨We built an AI system that turns research-level math into Lean and machine-checks it. We formalized 5 complex STOC papers to Lean, and even found a gap in one published proof. Paper: arxiv.org/abs/2606.31134. Thanks to @DARPA (expMath program) & @patrickshafto for their support.
6
647
Max Springer retweeted
New preprint! We introduce a new benchmark, SciConBench, with 9.11k scientific questions derived from Cochrane Systematic Reviews. We find evidence that frontier AI agents **cannot** synthesize scientific conclusions well. A thread 🧵 w/ @hayounggjung, @korolova & others
11
53
7
201
40,091
Incredibly sad to miss out on #ICLR2026 in Rio this week (and the churrasco!) but thankfully my fantastic co-authors will be there presenting several of our papers. Here is a full rundown, with info on each paper's presentation at the end /🧵
1
4
201
5/ 📄 Bi-Criteria Metric Distortion 4/25, 9am, Pavilion 4 [P4-#4205] arxiv.org/abs/2412.10671 📄 Less is More: Adaptive Coverage Sampling arxiv.org/abs/2504.14508 + 📄 SYNAPSE-G: Positive Mining from LLM Seeds arxiv.org/abs/2508.09544 DATA-FM Workshop, 4/26, 203 A/B
1
75
6/ 📄 Verifying Alignment Constraints in Composite-Data Regimes Workshop on Algorithmic Fairness Across Alignment Procedures & Agentic Systems, 4/26, 211 📄 Towards Statistical Verification for Trustworthy AI Workshop on Principled Design of Trustworthy AI, 4/27, 204 A/B
57
My 21st Ph.D. student, Suho Shin, just defended his thesis Game Theory & joins Stanford and MIT for postdocs. All 21 of my students’ theses focus on challenging deep theoretical foundations applied to auctions, games, big data, networks, algorithms, distributed & intelligent sys
29
91
3
1,517
73,086
Meet the Members of the Independent International Scientific Panel on AI. ✨ Forty experts from around the world coming together to put science at the core of international AI efforts. Learn more: tinyurl.com/ai-panel-members
3
26
2
57
10,592
Tuesday Seminars begin again for the semester tomorrow 1/27 with Max Springer's (@MaxSpringer23) talk entitled "A Geometric Perspective on LLM Behavior". citp.princeton.edu/events/20… Get here early for a sandwich & a chat. Seminar promptly at 12:15. Sherrerd 306. Livestream 🔗👇
1
1
2
171
Max Springer retweeted
Days in a work week: 5 Days in a month: 30 Total new submissions to arXiv in September: 26,646 arXiv editorial and user support staff: 7 someone who is good at science please help me with this. our team isn't sleeping. #openaccess #preprints
56
100
31
846
89,672
Max Springer retweeted
Max Springer grew up reading @sciam, but he never expected to write for the prestigious publication. Read how this recent @math_umd grad cast a media spotlight on math, including writing about the mathematics of @LoveIsland and promoting #PiDay on TV: go.umd.edu/springer
2
4
330
Max Springer retweeted
A team led by Mohammad Hajiaghayi (@MTHajiaghayi) has solved the Steiner Forest problem—a 35-year-old mathematical quandary. The resulting 126-page proof and algorithm could boost chip design for computing and AI, cutting energy use and costs. Read more: go.umd.edu/1zd2
4
20
1,456
Max Springer retweeted
After his @math_umd Ph.D., Max Springer is heading to @PrincetonCS to continue his #GameTheory research as a postdoc! While at #UMD, he interned at @nokiabell, Google Research and @NIH and was selected as an @aaas Mass Media Fellow and @NSF Graduate Research Fellow. #UMDGrad
1
3
340
Part of my current group of 20 exceptional PhD students, many already with medals in IOI, IMO, and ICPC from different parts of the world, focuses on Theory, likely the largest theory groups in the world. I am proud of all my current students and 17 already PhD graduated ones.
1
36
2,814