In a groundbreaking revelation that has left both hackers and ethical cybersecurity experts Googling ‘moral compass,’ researchers have discovered that prompt injection attacks on the Model Context Protocol (MCP) can allegedly be used for good. Or evil. Or something in between. Kind of depends on the time of day and your mood.

Think of MCP like your well-meaning but easily influenced friend who is susceptible to bad advice—but can also occasionally give you a heads-up about a pyramid scheme. It’s a paradox wrapped in a security blanket, and it doesn’t even come with a manual—because who needs those? Certainly not the AI industry.

Tenable, the new Michelangelo of digital art heists, insists that this Houdini-like versatility of MCP could be the Swiss army knife of our digital age. By deploying prompt injection, you could either defend the realm or conquer it. But hey, why not both? So next time you’re bored, just remember: You hold the keys to world peace. Or world domination. No pressure.