AI vs AI: How Anthropic's Claude Cracked OpenAI's Defenses in Under 72 Hours

Hackers breached OpenAI's internal systems using Anthropic's Claude AI model in under 72 hours, exploiting an image processing flaw in a community forum to compromise employee accounts and access private GitHub code. The white-hat team at Hacktron AI demonstrated that modern AI agents can chain vulnerabilities faster than human teams, compressing months of work into days. OpenAI paid a $6,500 bug bounty and patched the flaws, but the incident underscores a sobering reality:...