Anthropic warns of AI sabotage in the future, but current models are safe

From Cointelegraph
October 18, 2024 5:29:45 pm:

A firm’s investigation identified four “sabotage” threat vectors for AI, with “minimal mitigations” deemed adequate for current models.

Read more at Cointelegraph: Anthropic says AI could one day ‘sabotage’ humanity but it’s fine for now