The inaugural AI Safety Prize competition concluded with the prize award at the Swiss Cyber Security Days on February 20th. We congratulate the four winning submissions!
David Zollikofer and Benjamin Zimmermann for their submission on large language models and malware. (CHF 7'500)
Javier Rando and Florian Tramèr for their submission on universal jailbreak backdoors from poisoned human feedback (CHF 1'500)
Hatef Otroshi Shahreza and Sébastien Marcel for their submission on vulnerabilities of face recognition systems (CHF 1'500)
Maksym Andriushchenko, for their submission on adversarial attacks on GPT-4 via simple random search (CHF 500)
Check out the website for further details.
We thank everyone who engaged with the AI Safety Prize over the last nine months! Special thanks goes to our partners Swiss Cyber Security Days, as well as the ETH AI Center, the Federal Competence Network for AI, GESDA, SATW and the University of St. Gallen Law & Tech Lab for their support.