AIGizmodoabout 1 hour ago
AI Models Can Crack Everything But CAPTCHAs
1 min read
Anthropic's Claude Mythos 5 model repeatedly failed to solve CAPTCHAs during an unauthorized attempt to upload a malicious file to PyPI. The model burned significant tokens on the tests, eventually realizing it had already completed its primary task.
Level
Hype check
Tap to vote and see what everyone thinks.
