A new warning is circulating in security circles: the Model Context Protocol (MCP), an increasingly common way for AI systems to connect to data and tools, may not be ready for the age of quantum computing.
According to Security Boulevard, MCP deployments are vulnerable to what researchers call "Harvest Now, Decrypt Later" attacks. The idea is simple but unsettling. An attacker doesn't need a working quantum computer today. They can quietly copy encrypted data now — the kind of sensitive information flowing through AI infrastructure — and store it, betting that future quantum machines will be powerful enough to break today's encryption and unlock it later.
Security Boulevard frames the core question directly: is your MCP deployment leaking data, and how do you secure it against this emerging class of threats? The concern is that data considered safe today could become readable once quantum decryption becomes practical.
The worry extends beyond technology teams. According to theinsurer.com, the industry is now asking whether cyber insurers should be preparing for quantum computing now — a sign that the risk is being taken seriously not just by engineers but by the businesses that underwrite digital risk.
Together, the two reports point to a shift in how security is being discussed. Rather than treating quantum computing as a distant, abstract threat, researchers and insurers are urging organizations to think about protection today, before quantum machines arrive.
Why it matters: as AI systems like those built on MCP handle more sensitive data, the possibility that today's secrets could be unlocked by tomorrow's quantum computers means the time to strengthen defenses may be now, not later.