An AI cybersecurity expert details how Anthropic's Claude model behaved when prompted to find and patch bugs, demonstrating its capabilities for cyberdefense. When given insecure code, the model initially refused a direct security review but complied when asked to fix the code, indicating its intended function for security applications.
Opening Kapyn…