Jailbreak Anthropic’s new AI safety system for a $15,000 reward

/in

In testing, the technique helped Claude block 95% of jailbreak attempts. But the process still needs more ‘real-world’ red-teaming.

Latest stories for ZDNET in Security – ​Read More

Cybercriminals Court Traitorous Insiders via Ransom NotesCybercriminals Court Traitorous Insiders via Ransom NotesChinese ‘Infrastructure Laundering’ Abuses AWS, Microsoft CloudChinese ‘Infrastructure Laundering’ Abuses AWS, Microsoft Cloud