Taylor Monahan Says Crypto Teams Need Monitoring and Circuit Breakers, Not More Audits
Summary: Security expert Taylor Monahan advises crypto projects to shift focus from repeated audits to real-time monitoring and automatic circuit breakers. She estimates audits cover 90% of risk, leaving a residual 5% that requires autonomous shutoffs. Her comments follow recent exploits at K3 Capital’s sBOLD vault and perpetual exchange Lighter.
Security expert Taylor Monahan has advised cryptocurrency development teams to prioritize real-time monitoring and automatic circuit breakers over repeated security audits, marking a significant shift in her guidance over the past six months. Speaking on the Nov. 18 episode of Unchained’s Uneasy Money, which she co-hosts, Monahan stated that while her previous counsel was “more classic advice,” her current recommendation is focused heavily on visibility.
“Right now my advice is, like, monitoring,” Monahan said. “You need to have visibility, and you need to just crank that through the roof.” She attributed this change in strategy to the increasing prevalence of AI agents probing projects for vulnerabilities and the ability of AI tools to help smaller teams detect anomalies that previously required large, highly skilled staff. “The AI is very good at leveling your own team up,” she noted, adding that “the next six months are gonna be wild in this regard.”
The Limits of Audits and Formal Verification
Monahan provided a quantitative estimate of the protection offered by traditional security measures. In her view, “the audit gets us 90%” of risk mitigation. She added that formal verification, a method that uses mathematics to prove code behaves as designed, contributes an additional 5%. However, she warned that “you still have 5% that it might just completely collapse.”
To address this residual risk, Monahan advocated for the implementation of a circuit breaker layer. This mechanism allows a team to define specific conditions that must never occur, such as a large unauthorized outflow of funds. If such a condition is met, the system automatically halts operations. “It’s just gonna pull the plug,” she explained. As a sample rule, she suggested: “Under no circumstances should $200 million ever leave this contract.”
Monahan emphasized that if an attack is detected immediately and reacted to autonomously, the protocol can survive. “Instead of being wrecked 100% of the way, you can be wrecked 5% of the way,” she said, describing circuit breakers as “that layer that historically in crypto we just have not had.” Host Kain Warwick supported the assessment, noting that while “audits catch a lot of stuff,” they “never catch the thing that kills you.”
Recent Incidents Highlight Vulnerabilities
Monahan’s comments arrived during a week marked by notable security incidents in the crypto sector. On Wednesday, K3 Capital said it had paused the contracts behind sBOLD, its yield vault built on the Liquity V2 lending protocol. The pause followed an alert from security firm Blockaid, which reported that attackers had drained approximately $3.43 million. Blockaid described the exploit as ongoing at the time of its initial alert, indicating that losses could potentially grow.
Earlier in the week, on Monday, the perpetual futures exchange Lighter froze withdrawals after funds were drained through what the platform identified as a trader’s leaked application programming interface (API) key. Lighter later stated that a contract safeguard known as Guardian had detected the anomaly. The exchange’s Security Council subsequently returned about $5 million in USDC to the affected trader, and withdrawals have since reopened.
AI in Cyberattacks
Monahan also addressed the role of artificial intelligence in cyberattacks, specifically noting its use by hackers tied to North Korea. She observed that these actors are currently using AI for minor tasks, such as having it “adjust their social engineering scripts.” While she described current usage as “so boring” and noted that “we’re not seeing the attackers optimize the AI yet,” she expects this to change as younger hackers begin building campaigns around AI from scratch. This follows her earlier observation in September that OpenAI still could not monitor its agents effectively.