Quantum risk to AI systems is the possibility that future quantum computers will weaken or break cryptographic methods that protect AI data, models, infrastructure and communications. The main organisational challenge is planning the move to quantum-resistant cryptography before sensitive information or services become exposed.

Quantum computers use different principles from conventional computers and may eventually solve certain mathematical problems much faster. Many widely used public-key cryptographic methods rely on problems that a sufficiently capable quantum computer could solve efficiently. The exact timing and practical capability are uncertain, but long-lived information can be at risk before such a machine exists.

Harvest now, decrypt later

In a harvest-now-decrypt-later attack, an adversary collects encrypted information today with the intention of decrypting it in the future. This matters for data that must remain confidential for many years, including personal, commercial, government or research information used to build or operate AI systems. AI services also rely on cryptographic protections for software updates, identity, model distribution, API connections and cloud infrastructure.

Q-Day is an informal term for the point at which a cryptographically relevant quantum computer can break widely used public-key systems. There is no single confirmed date. Planning against a fixed prediction can be misleading because organisations may need years to discover where cryptography is used, select replacements and update connected systems.

Build crypto-agility

Crypto-agility is the ability to replace cryptographic algorithms and protocols without rebuilding an entire service. Organisations can begin by inventorying where cryptography protects AI data and operations, including supplier services, stored datasets, model registries, devices, certificates and backups. They should identify information with a long confidentiality life and systems that would be difficult to update quickly.

Migration planning should account for dependencies across vendors and infrastructure. Newer post-quantum cryptographic standards are intended to resist known quantum attacks, but selection and deployment need to follow authoritative standards and security guidance. Hybrid approaches may be used during transition, where appropriate, while organisations test compatibility, performance and fallback procedures.

For AI governance, the issue is both technical and strategic. A model may remain accurate while the confidentiality or integrity of its data and delivery chain is undermined. Assign an owner for the transition, include cryptographic dependencies in supplier reviews and rehearse how critical AI services would continue if a protection method had to be replaced urgently. Early inventory and staged change reduce the risk of a rushed response later.