Ledger Urges ‘Human Sovereignty’ Security Model as AI and Quantum Risks Rise

Ledger, one of the longest-running names in hardware wallets, is urging Korea’s financial sector to rethink what “security” means in a digital-asset era—arguing that the last line of defense is not a feature or a policy, but 'human sovereignty' built into system architecture.

Speaking Tuesday at the Institutional Web3 Forum in Seoul, Takatoshi Shibayama, APAC head of Ledger Enterprise, said the rapid convergence of traditional finance and digital-asset infrastructure under emerging regulation is forcing institutions to reassess assumptions that worked for legacy systems. “This is the moment that determines whether Korea remains a country that creates wealth—or a country that preserves wealth,” he said, adding that protecting assets is ultimately harder than generating them.

The event, held at the Glad Hotel in Seoul’s Yeongdeungpo district, brought together roughly 100 invited participants from banks, brokerages, insurers, fintech firms, and digital-asset companies. It was co-hosted by TokenPost, the Korea Fintech Industry Association (KORFIN), and the Open Blockchain & AI Association (OBDIA), with Bithumb, Coinone, and Korbit serving as official sponsors.

Shibayama framed his remarks around a central distinction: in traditional financial breaches, attackers often steal data; in crypto, they can steal 'the money itself'. “You can’t apply the same security standards,” he said, calling for a shift from software-centric controls to architecture-level protection that keeps final authorization in human hands.

In his presentation, titled “The paradox of technological progress: why human sovereignty is the final line of defense,” Shibayama outlined three accelerating threat vectors—AI, governance attacks, and quantum computing—arguing that they differ in form but converge on a shared failure mode: 'blind signing', where users authorize transactions without being able to independently verify what is being approved.

AI: always-on adversaries and automated crime

Shibayama warned that AI agents are rapidly expanding what both legitimate users and criminals can automate—ranging from code generation and contract negotiation to convincing human impersonation. As trading and treasury workflows shift toward AI-driven execution, he said, criminal operations will follow. “AI can identify infrastructure vulnerabilities and run real-time attacks 24/7,” he cautioned.

His proposed mitigation emphasized hardware-enforced separation of duties. Instead of relying on patches and software prompts alone, Shibayama argued institutions should adopt a model where software merely requests permissions, a human must approve, and the final signing key is physically separated from AI authority. The principle, he said, is to ensure that automation cannot silently escalate into unilateral control over funds.

Governance attacks: when the protocol works—on the wrong input

The second risk area focused on governance and authorization pathways—often treated as operational plumbing rather than primary attack surfaces. Shibayama cited incidents involving Drift and Kalda to illustrate how attackers can extract large sums by manipulating the decision-making layer rather than “breaking” cryptography.

In the Drift case, he said social engineering enabled attackers to steal $285 million in about 12 minutes. In the Kalda incident, he described a different mechanism: malicious instructions were inserted into validator-node processes so that funds were routed to attacker-controlled addresses. Despite differing methods, he argued the common denominator was that the system dutifully executed valid actions based on compromised or misleading inputs.

“The protocol operated normally, but it received incorrect information,” Shibayama said, adding that the real breakdown occurred because the authorizing party—often an automated workflow—could not clearly see what it was signing. For institutions, he framed 'human-readable verification' not as a premium feature but as a baseline requirement for any transaction approval flow.

Quantum computing: shorter timelines, higher urgency

The third threat was quantum computing, which Shibayama said is increasingly being discussed not as a distant risk but as a medium-term planning constraint. Citing a March white paper from Google’s AI division, he said the resources required to break the elliptic-curve cryptography used by Bitcoin (BTC) and Ethereum (ETH) have been estimated at roughly 20 times less than previously believed, with the paper pointing to around 500,000 qubits as a reference level.

He noted that Google’s “Willow” system is currently at 105 qubits, while the company is targeting a post-quantum transition by 2029. By comparison, he said, banks and governments often work on 2030–2035 migration schedules—timelines that may be too slow for digital assets given the irreversibility of onchain transfers and the scale of automated settlement. “Digital assets need to respond much faster,” he said.

From “features” to architecture

Across the three threat categories, Shibayama returned to the idea that modern crypto risk is increasingly about authorization—who can approve what, and under what visibility—rather than only perimeter defense. In his view, 'blind signing' is the core systemic weakness: users believe they are in control, but in practice they approve actions without sufficient verification.

Ledger’s recommended direction centers on defaults: hardware-based security, explicit signing structures that make transaction intent readable to humans, and post-quantum-ready architecture planning. “No transaction should be executed if a human cannot verify what it is,” Shibayama said.

He concluded by rejecting the framing of security as a discretionary spend or a reactive add-on. “Security is not a cost, not after-the-fact response, and not a feature,” he said. “The final line of defense is human sovereignty.”