Joshua Lamerton
Post-quantum cryptography is often presented as a certificate-upgrade project. That framing is too narrow for agentic systems.
AI agents create large numbers of machine identities, short-lived credentials, delegated permissions, signed tool calls, and durable evidence. These systems are being deployed while organizations are also preparing to replace cryptographic algorithms vulnerable to future quantum computers.
The two transitions interact.
A conventional application may authenticate a user and call a small set of services. An agent can dynamically discover tools, delegate work to other agents, operate across organizations, and continue a task long after the initial human interaction.
Each step may rely on cryptography for:
If those dependencies are not inventoried, post-quantum migration will miss critical paths.
Sensitive traffic captured today may be decrypted in the future if it remains valuable long enough. That is the familiar confidentiality concern.
Agent systems add an integrity concern. Evidence may need to remain trustworthy for years: who authorized an action, which model and policy were active, and whether a record was altered. A signature that becomes forgeable can weaken the historical assurance of the system even if the underlying data is no longer secret.
This is particularly important in finance, healthcare, infrastructure, and autonomous transactions.
The objective is not to replace every algorithm immediately. It is to make replacement possible without rebuilding the entire agent platform.
That requires crypto-agility:
Agent protocols should negotiate supported capabilities without silently downgrading security.
Research published in 2026 specifically examines post-quantum migration for agentic systems, including MCP and agent-to-agent communication. These boundaries deserve attention because they can connect independently operated systems with different migration timelines.
A secure design should bind identity, authorization, and protocol context into cryptographic evidence. It should also distinguish the human principal from the acting agent and any delegated sub-agent.
Organizations should begin with discovery. Identify data that requires long-term confidentiality, evidence that requires long-term integrity, and external dependencies that control migration timing. Prioritize high-value agent workflows and introduce algorithm versioning before a forced transition.
The strategic lesson is broader: agent infrastructure needs cryptography as an adaptable service, not a hidden library choice.
Post-quantum readiness is therefore not only about future quantum machines. It is about whether today’s rapidly expanding autonomous systems can survive a fundamental change in the trust layer beneath them.