Anthropic security framework auto-downgrades: Claude mistakenly deletes developer's 700 GB home directory; automated cleanup becomes a disaster
When a developer tested AI file deletion protection, the Anthropic safety framework automatically downgraded the model, and an error in reusing variable names during the post-test cleanup step mistakenly deleted the entire user home directory. Although some data was recovered, losses remained, exposing a new AI safety problem.
Return to original text
A short introduction is retained here; the full text from the source website is not reposted. For specific wording and subsequent updates, please refer to the original article.
Open the original article on ITHome