Attack report shows the limits of bad intentions: Claude blocks attackers
Anthropic published a report on how various actors are trying to misuse the Claude model for nefarious purposes. Most attempts apparently fail or are disrupted by Anthropic, suggesting that current security barriers are holding up for now.