# Vasilios Mavroudis

> Machine Learning Researcher

**Wikidata**: [Q130968194](https://www.wikidata.org/wiki/Q130968194)  
**Source**: https://4ort.xyz/entity/vasilios-mavroudis

## Summary
Vasilios Mavroudis is a British computer scientist and artificial intelligence researcher specializing in systems security and AI safety. He currently serves as a principal investigator at the Alan Turing Institute, where he conducts research on adversarial machine learning, cybersecurity, and the security implications of emerging technologies. He has contributed to major initiatives including the International AI Safety Report and has presented his findings at premier security conferences such as Black Hat, DEF CON, and RSA Conference.

## Biography
- Born: 1989
- Nationality: United Kingdom
- Education: University College London
- Known for: Research on AI safety, systems security, adversarial attacks on machine learning systems, and ultrasound ecosystem privacy
- Employer(s): Alan Turing Institute (current)
- Field(s): Artificial Intelligence, Systems Security, Computer Security

## Contributions
Vasilios Mavroudis has made significant contributions to cybersecurity and artificial intelligence safety through his research on vulnerabilities in emerging technologies. His notable works include "On the Privacy and Security of the Ultrasound Ecosystem," which examined privacy risks in ultrasonic cross-device tracking systems. He presented this research at Black Hat Europe 2016 in a briefing titled "Talking Behind Your Back: Attacks and Countermeasures of Ultrasonic Cross Device Tracking."

In 2025, he contributed to the "International AI Safety Report," a government-commissioned document addressing global AI safety concerns. Additionally, he co-authored "Humanity's Last Exam" (arXiv:2501.14249), a research paper focused on evaluating AI systems. His research has also explored deep backdoors in deep reinforcement learning agents, which he presented at Black Hat USA 2024.

Mavroudis has participated in major cybersecurity conferences including DEF CON 25, Black Hat USA (2017, 2024), and the RSA Conference. His work spans theoretical contributions, such as "Online Convex Optimisation: The Optimal Switching Regret for all Segmentations Simultaneously," to practical security analyses. He has been recognized by the Heidelberg Laureate Forum and has provided expert input to the Parliament of the United Kingdom on AI-related matters in 2024. His research has been covered by media outlets including ZDFinfo, Tech Monitor, and Spektrum der Wissenschaft.

## FAQs

### Q: What is Vasilios Mavroudis's current role?
A: Vasilios Mavroudis is currently a principal investigator at the Alan Turing Institute, the UK's national institute for data science and artificial intelligence, where he conducts research on AI safety and cybersecurity.

### Q: What are Vasilios Mavroudis's main research areas?
A: His primary research areas include artificial intelligence safety, systems security, computer security, adversarial machine learning, and privacy issues in emerging technologies such as ultrasonic tracking systems.

### Q: What conferences has Vasilios Mavroudis presented at?
A: He has presented at major cybersecurity conferences including Black Hat USA (2017, 2024), Black Hat Europe (2016), DEF CON 25, and the RSA Conference.

### Q: What is "Humanity's Last Exam"?
A: "Humanity's Last Exam" is a research paper co-authored by Vasilios Mavroudis (arXiv:2501.14249) that focuses on evaluating AI systems and their capabilities.

### Q: Has Vasilios Mavroudis contributed to government policy?
A: Yes, he contributed to the International AI Safety Report published by the UK government in 2025 and has participated in UK Parliament committee discussions on AI-related matters.

## Why They Matter
Vasilios Mavroudis represents a critical bridge between academic AI research and practical security implementation at a time when AI systems are becoming increasingly integrated into society. His work addresses fundamental questions about AI safety and security with far-reaching implications for technology deployment. As a contributor to the "International AI Safety Report" and through his engagement with the UK Parliament, he has influenced policy discussions around AI governance at the national level.

His research on ultrasonic tracking and deep reinforcement learning backdoors has exposed previously unrecognized vulnerabilities, prompting industry-wide reassessments of security practices. By identifying how AI systems can be exploited or manipulated, Mavroudis's work helps establish frameworks for developing more robust and trustworthy AI technologies. His presence at premier security conferences like Black Hat and DEF CON ensures that his findings reach practitioners who can implement protective measures. As AI systems become more prevalent, researchers like Mavroudis who understand both the potential and the vulnerabilities of these technologies play an essential role in ensuring their safe deployment across industries.

## Notable For
- Principal Investigator at the Alan Turing Institute, the UK's national institute for data science and AI
- Contributor to the "International AI Safety Report" (2025), a major government-commissioned AI safety document
- Multi-year speaker at Black Hat conferences (2016, 2017, 2024) addressing security vulnerabilities
- Recognized by the Heidelberg Laureate Forum for contributions to computer science and data privacy research
- Expert contributor to UK Parliament committees on AI-related matters (2024)

## Body

### Early Life and Education
Vasilios Mavroudis was born in 1989 and holds British citizenship. He pursued his education at University College London, a constituent college of the University of London federal research university. The University of London, established in 1836, is headquartered at Senate House on Malet Street in London.

### Research Career
Mavroudis has established himself as a researcher at the intersection of artificial intelligence and security. His work spans multiple domains within computer science, including artificial intelligence safety, systems security, and adversarial machine learning. He is affiliated with the Alan Turing Institute, a research institute founded in 2015 and headquartered at the British Library in London.

He maintains an active research profile, documented through his Google Scholar author ID (b10w74sAAAAJ), DBLP author ID (154/8074), and Scopus author ID (57200514266).

### Key Publications and Presentations

**Academic Papers and Reports:**
- "On the Privacy and Security of the Ultrasound Ecosystem" — Investigating privacy risks in ultrasonic technologies
- "Humanity's Last Exam" (arXiv:2501.14249) — Research on AI system evaluation
- "Online Convex Optimisation: The Optimal Switching Regret for all Segmentations Simultaneously"
- "International AI Safety Report" (2025) — Government-commissioned report on AI safety
- "A Touch of Evil"

**Conference Presentations:**
- Black Hat USA 2024: "Deep Backdoors in Deep Reinforcement Learning Agents"
- Black Hat USA 2017: Security research presentations
- Black Hat Europe 2016: "Talking Behind Your Back: Attacks and Countermeasures of Ultrasonic Cross Device Tracking"
- DEF CON 25: Security research presentations
- RSA Conference: Expert presentations on AI security topics

### Professional Recognition
Mavroudis has received recognition from the Heidelberg Laureate Forum, which highlighted his work balancing verifiability and privacy in data systems. His expertise has been sought by governmental bodies, including the Parliament of the United Kingdom, where he participated in committee discussions on April 24, 2024.

### Media and Public Engagement
His work has been featured in various media outlets, including Spektrum der Wissenschaft (Scientific American's German edition), ZDFinfo, and Tech Monitor, demonstrating his ability to communicate complex technical concepts to broader audiences. He also has an IMDb profile (nm10084646).

### Online Presence
- Website: https://mavroud.is
- Twitter/X: @mavroudisv (active since 2011)
- GitHub: https://github.com/mavroudisv/

## Schema Markup
```json
{
  "@context": "https://schema.org",
  "@type": "Person",
  "name": "Vasilios Mavroudis",
  "alternateName": ["Vasilis Mavroudis", "Vas Mavroudis"],
  "jobTitle": "Principal Investigator",
  "worksFor": {
    "@type": "Organization",
    "name": "Alan Turing Institute"
  },
  "nationality": {
    "@type": "Country",
    "name": "United Kingdom"
  },
  "birthDate": "1989",
  "alumniOf": {
    "@type": "EducationalOrganization",
    "name": "University College London"
  },
  "knowsAbout": [
    "Artificial Intelligence",
    "Systems Security",
    "Computer Security",
    "Machine Learning",
    "AI Safety"
  ],
  "description": "British computer scientist and AI safety researcher at the Alan Turing Institute specializing in systems security and adversarial machine learning",
  "sameAs": [
    "https://mavroud.is",
    "https://x.com/mavroudisv",
    "https://github.com/mavroudisv/",
    "https://scholar.google.com/citations?user=b10w74sAAAAJ",
    "https://www.imdb.com/name/nm10084646/"
  ],
  "award": "Heidelberg Laureate Forum"
}

## References

1. [Source](http://www0.cs.ucl.ac.uk/staff/V.Mavroudis/)
2. [Source](https://www.turing.ac.uk/people/researchers/vasilios-mavroudis)
3. [Source](https://scilogs.spektrum.de/hlf/10-out-of-200-between-verifiability-and-privacy-vasilios-mavroudis-makes-sense-of-our-data/)
4. [Source](https://www.imdb.com/name/nm10084646/)
5. [Source](https://www.gov.uk/government/publications/international-ai-safety-report-2025)
6. [Source](https://arxiv.org/abs/2501.14249)
7. [Source](https://committees.parliament.uk/event/21367)
8. [Source](https://www.imdb.com/title/tt8232662/?ref_=nm_knf_t_1)
9. [Source](https://www.techmonitor.ai/digital-economy/ai-and-automation/agentic-ai-cybersecurity-implications)
10. [Source](https://www.youtube.com/watch?v=WW1-xnTIDjQ&t=8s)
11. [Source](https://www.blackhat.com/us-24/briefings/schedule/#deep-backdoors-in-deep-reinforcement-learning-agents-39550)
12. [Source](https://www.blackhat.com/eu-16/briefings.html#talking-behind-your-back-attacks-and-countermeasures-of-ultrasonic-cross-device-tracking)
13. [Source](https://www.blackhat.com/us-17/briefings.html#vasilios-mavroudis)
14. [Source](https://www.defcon.org/html/defcon-25/dc-25-speakers.html#Mavroudis)
15. [Source](https://www.youtube.com/watch?v=UJFQ_i0uyns)
16. [Google Scholar](https://scholar.google.com/citations?user=b10w74sAAAAJ&hl=en&oi=sra)
17. [Source](https://x.com/mavroudisv)