Three testing models with the same goal but different directives engaged in "increasingly aggressive" territorial attacks on one another, according to Anthropic.π Read more.
π Via "Dark Reading"
----------
ποΈ Seen on @cibsecurity
π‘ π‘ Cybersecurity & Privacy π‘ - News @cibsecurity Β· 28.5K subscribers Three testing models with the same goal but different directives engaged in "increasingly aggressive" territorial attacks on one another, according to Anthropic.