Loading ...Claude Hacked Three Real Organizations During Botched Test
Anthropic’s efforts to test Claude’s offensive cybersecurity skills produced an unintended real-world result: Its AI models gained unauthorized access to three outside organizations.
https://www.zerohedge.com/ai/claude-hacked-three-real-organizations-during-botched-test
I used Claude last night to put together a recovery program for a small group of apple trees over the next two years, complete with a spreadsheet for which chemicals to spray and when, when to prune, when to fertilize, how to grow bigger apples. All in about 10 minutes. Claude identified the diseases infecting the trees and the apples, how to prevent them going forward, and tree care during the winter.
Would have taken me hours to parse that out myself…