Advisory · 23 July 2026
AI cybersecurity risk in Web3
Development, operations, and third-party exposure.
AI does not need your private keys to move your assets. Zanarc's 28-page advisory examines how AI use across Web3 development and live operations creates risk that conventional enterprise controls do not address, and the controls platforms and their supervisors should expect.
Independently reviewed. The advisory passed two external accuracy reviews before release.
Standards-led. Zanarc leads the C4 AI Security Requirements for Cryptocurrency Systems working group, and the advisory maps its risks to the draft standard's seven control areas.
Evidence-based. The risks and controls draw on documented incidents and published research, referenced throughout.
Register your email address to download the advisory.
Register to download
What the advisory covers
Written for platform operators, custodians, infrastructure providers, and the supervisors and auditors who assess them. Inside the advisory:
Development
AI-generated code in security-critical systems, hallucinated dependencies and slopsquatting, and defects locked into non-upgradeable deployments.
Operations
AI agents in transaction and asset management, manipulated transaction requests, on-chain protocol interaction, and multi-agent exposure.
Third parties
Compromised models and agent frameworks, external model providers, poisoned retrieval sources, and persistent agent memory.
Why Web3 is different
Finality and recoverability, key material exposure, immutable code, composability, and the limits of human oversight.
Incident analysis
Lessons from the Bybit incident, the Base network prompt injection, and the July 2026 OpenAI model evaluation breach.
A targeted set of requirements
The seven control areas in the draft C4 AI Security Requirements for Cryptocurrency Systems standard, and how they map to the three domains.
Frequently asked questions
About the advisory, the registration process, and Zanarc.
Who is the advisory written for?
Platform operators and the development and security teams inside them, custodians and infrastructure providers, and the regulators and auditors who assess them. Section 4 summarises the control areas most relevant to supervisors and auditors.
Is the advisory free?
Yes. Register your email address and your copy of the 28-page PDF will be sent to your inbox.
How will my email address be used?
To deliver the advisory and, where you have opted in, occasional updates from Zanarc on digital asset cybersecurity. You can unsubscribe at any time.
What is the C4 AI Security Requirements for Cryptocurrency Systems working group?
A working group under the CryptoCurrency Certification Consortium (C4) developing a draft standard for the security of AI in digital asset infrastructure. Zanarc leads the working group, and the current draft is available from Zanarc on request.
Does the advisory replace NIST, ISO/IEC, or OWASP guidance?
No. General AI risk-management and application-security frameworks provide the governance foundation and platforms should adopt them. The advisory covers the cryptocurrency-specific conditions they do not prescribe controls for: cryptographic signing boundaries, final on-chain execution, permissionless protocol interaction, asset-specific recovery functions, and cross-chain state changes.
Who is Zanarc?
Zanarc is a specialist blockchain and digital asset cybersecurity consultancy. It provides cybersecurity and regulatory advisory to virtual asset platforms and their supervisors, CryptoCurrency Security Standard (CCSS) readiness assessments and audits, and AI cybersecurity risk assessment for cryptocurrency systems.
Register to download the advisory
An AI error in Web3 can become a final on-chain transaction. The advisory sets out why the controls must act before signing, and what platforms and their supervisors should put in place.