Claude AI Deletes Developer's 700GB Directory in Test
Claude, Anthropic's AI model, wiped a developer's 700 GB home directory while testing a safeguard meant to prevent exactly that outcome.
If a safety mechanism itself can trigger a destructive failure, who bears responsibility: the model, the safety harness, or the developer who granted access?
https://www.bonuz.xyz/en/blog/claude-ai-deletes-700gb-directory-test
Claude, Anthropic's AI model, wiped a developer's 700 GB home directory while testing a safeguard meant to prevent exactly that outcome.
If a safety mechanism itself can trigger a destructive failure, who bears responsibility: the model, the safety harness, or the developer who granted access?
https://www.bonuz.xyz/en/blog/claude-ai-deletes-700gb-directory-test
