Reposted by @larry-yang.bsky.social
NEW FROM @davidsirota.com: Maybe the scariest thing about AI is how well it learned from us.
The rogue AI agents in the Hugging Face hack colluded to evade rules and hide info from their overseers.
OpenAI did something similar in its own incident investigation.
www.levernews.com/the-machines...
levernews.com
The Machines Escaped. Their Masters Did So First.
The most frightening thing about rogue AI agents may be how faithfully they mirror the powerful humans who created them.