Our focus in this research was the use of data poisoning to insert ‘backdoors’ into llms. It’s trivially easy to poison llms into spitting out gibberish, says anthropic github copilot ‘camoleak’ ai attack exfiltrates data take note Tests of large language models reveal that they can behave in deceptive and potentially harmful ways
Cric Leaks (@cricleaks_) • Threads, Say more
What does this mean for the future?
