Twitxy Saved

Lukasz Olejnik

Verified account This account is verified. Learn more

@lukOlejnik

Security & Privacy. Data Protection. Research & Development. Engineering. Analyst. Policy. W3C. Consultant. Author. me@lukaszolejnik.com
Ph.D, LL.M. @warstudies
Brussels / London / EU lukaszolejnik.com Joined March 2013
267 Following 33K Followers 21K Posts

Latest posts 5

Pinned post

Lukasz Olejnik Verified account @lukOlejnik
I'm open to consulting, full-time or part-time work across research, engineering, security, AI, regulation and policy. AI-first work very much in scope. This can include commissioned research, adversarial security/privacy review, AI and agent-system, technical analysis, and policy/regulatory work. me@lukaszolejnik.com
Lukasz Olejnik Verified account @lukOlejnik Aug 27
I'm proud of my recent work. 1) AI-powered autonomous influence operations, 2) AI-assisted cryptanalysis, and 3) a concrete break of a proposed FHE construction.

- IO Factory - an agentic environment for simulating AI-enabled influence campaigns at scale, including runs with up to 100,000 agents. I learned a lot here about building AI-powered systems, orchestration, simulation, and using AI coding and testing assistants while keeping experiments traceable and reproducible.

URL: arxiv.org/abs/2608.10920

- AI Grinding for Fun and Cryptanalysis - a framework for automated AI-assisted cryptanalysis. Agents generate, test, and refine attack hypotheses at scale. The workflow uncovered multiple security failures in published cryptographic constructions.
arxiv.org/abs/2608.21986

- Key Recovery from Residue-Confined Errors in the CRT-RLWE Construction - focused cryptanalysis of a proposed FHE construction. eprint.iacr.org/2026/1734
View quoted post on X
3 replies 10 reposts 44 likes 18K views View on X
Lukasz Olejnik Verified account @lukOlejnik
An object of unknown origin, possibly a missile fell in southeastern Poland this morning causing a loud explosion. Authorities investigating an airspace violation amid Russian attacks on Ukraine. No injuries reported.
Lukasz Olejnik Verified account @lukOlejnik
Someone hacked Prime Intellect's inference servers and replaced DeepSeek V4.1 Flash with an abliterated version stripped of safety refusals. For 53 hours "safety" refusal rates collapsed to ~1% vs 83-98% originally.
4 replies 3 reposts 28 likes 1.9K views View on X
Lukasz Olejnik Verified account @lukOlejnik
AI-powered cyberattacks on European Commission and EU systems using ARTEX, a framework for autonomous AI-driven offensive action. Affected internal administrative services, including translation and reimbursement tools, databases used for salaries and timesheets. These cyberattacks are unrelated to the EU AI Act.