Xander Davies

Anthropic's joint study reveals that as few as 250 malicious documents can create backdoor vulnerabilities in large language models regardless of model size, challenging assumptions that attackers need to control a percentage of training data.

“Author of "Anthropic Battles Poisoned Data, Branding Blitz, and Legal Bills" in Anthropic”

“Researcher who publicly disclosed developing a jailbreak that extracted malicious responses from Fable 5 within hours of access.”