Ledger CTO Warns AI Is Lowering Attack Costs and Worsening Crypto Security Threats

As artificial intelligence becomes increasingly integrated into Web3, the automated management of crypto assets by AI agents is gaining traction, while creating serious risks of manipulation by hackers. Ledger, a leading crypto hardware-wallet maker, estimates that it protects more than 20% of the world's crypto assets. A central challenge for decentralized-finance security is enabling AI agents to interact smoothly with wallets without compromising private keys or allowing unauthorized fund transfers.

Ledger launched a new open-source toolkit called Agent Stack on June 10, 2026, allowing AI agents to interact with wallets. Built around a model in which agents propose and humans approve, the toolkit lets AI agents check balances and draft transactions. Critical fund transfers and signatures must still be confirmed by users on a physical Ledger device, helping prevent hackers from stealing funds.

Decentralized finance, or DeFi, relies on smart contracts and cross-chain bridges to manage assets, and exploited configuration errors or software vulnerabilities can quickly cause massive losses. Anthropic’s next-generation Claude Fable 5 model has stronger reasoning and coding capabilities. While that can make developers more productive, it could also lower the barriers for hackers seeking to analyze code and build attack tools.

As of July 20, 2026, incident data showed that DeFi losses from hacks had exceeded $840 million this year. Security experts said Claude Fable 5 might not devise entirely new attack methods, but it could identify configuration mistakes, locate vulnerabilities and construct exploits at near-superhuman speed, further compressing cybersecurity teams’ response time.

Cryptocurrency platforms have long faced risks from smart-contract and system vulnerabilities, and generative AI is rapidly lowering the barriers for attackers to find, analyze and exploit weaknesses. Mitchell Amador, CEO of blockchain security platform Immunefi, said the proliferation of new models such as Claude Opus 4.8 and ChatGPT 5.5 has fueled a “vulnerability apocalypse” in crypto security.

Amador said AI models are a leading cause of the surge in cryptocurrency hacks and losses in 2026. More than $634 million was stolen from crypto platforms in April 2026 alone, marking both a sharp increase in monthly security losses and the highest level since early 2025. The figures underscore the need for operators to accelerate improvements in vulnerability detection and defenses.

HashiCorp co-founder Mitchell Hashimoto said AI-assisted development can accelerate the launch of blockchain projects, but generated code has a higher defect rate than code written by humans. Such systemic risks are challenging the crypto industry’s “Code is Law” security foundation, which relies on immutable code to preserve trust.

The latest research found that blockchain attacks stemming from vulnerabilities in AI-generated code rose 30% in the six months before the report was published. The report did not identify the research organization, the exact statistical period or the losses involved. Hashimoto warned that faster development could also widen the attack surface without human review and security audits.

Blockchain security company CertiK said the cryptocurrency industry faces AI-enhanced social engineering and attacks exploiting protocol vulnerabilities. North Korean hackers are also using generative AI to fabricate identities and make scams more convincing, exposing exchanges, DeFi protocols and investors' assets to increasingly complex risks across platforms.

A CertiK report showed that cryptocurrency hacks have caused more than $600 million in losses so far in 2026. It predicted that deepfakes, phishing and software supply-chain compromises will drive the year's biggest attacks. The company urged investors to improve their awareness of offline storage, multifactor authentication and transaction verification.