Analyst warns AI fraud outpaces crypto defenses

TLDR

An influential crypto analyst is warning that AI driven fraud is scaling faster than the industrys current security defenses.

1. Chain analysis and industry data show AI powered scams are more profitable and efficient than older crypto fraud methods.
2. Major exchanges and regulators are racing to deploy their own AI defenses, but experts see only a short window before attackers catch up.
3. For everyday users and builders, the biggest new risks involve deepfakes, impersonation, and over trusting AI agents that can move funds.

Deep Dive

1. How AI Is Supercharging Crypto Fraud

Analyst Ali Martinez cites research showing AI enabled scams were about 4.5 times more profitable than traditional scams in 2025, with AI driven fraud totaling roughly $17 billion and some models achieving success rates above 70 percent. These scams increasingly use deepfake video, voice, and AI generated messages to create convincing romance and investment schemes, making social engineering much harder to spot than older copy paste phishing attempts.

AI is also starting to attack crypto specific tooling. A recent exploit against a Grok connected Bankr wallet on Base used a prompt injection delivered via a seemingly harmless NFT, tricking an AI agent into authorizing transfers without any key compromise, exposing a new class of agent based risk.

2. How Crypto Defenses Are Responding

Martinez describes an emerging AI vs AI arms race, where large platforms deploy AI to detect and block AI powered attacks. Binance, for example, reports that more than half of its fraud controls now use AI and says these systems helped block over $10.5 billion in potential user losses and intercepted 22.9 million scam or phishing attempts in a recent 15 month window.

Regulators are also taking AI specific cyber risk seriously. The Bank of Spain has warned that powerful vulnerability spotting models such as Anthropics Mythos could shorten the time from bug discovery to exploitation, and it is calling for broader access to defensive models like Glasswing and tighter international coordination on cyber resilience.

Security researchers highlight that cutting edge models already outperform earlier forecasts on offensive cyber tasks, which suggests defenders have only a narrow window of a few months to integrate similar capabilities before attackers do.

3. Practical Implications For Users And Builders

For individual crypto users, the biggest near term risks are upgraded versions of familiar scams. Deepfake executive streams, fake airdrops, and impersonation tokens piggyback on trusted brands, while AI generated support chats or DMs try to nudge you into connecting a wallet or revealing recovery information. Treat any too good to be true offer, even if it looks highly polished, as untrusted until verified through official channels.

For builders, the new failure mode is over empowering AI. Systems where an AI agent can initiate transfers or approve contract interactions should separate analysis from execution, require human or multi factor confirmation for meaningful movements, and hard limit what an agent can do with any single prompt. Security teams will increasingly need AI for anomaly detection and code review, but it must be layered with traditional controls, audits, and clear permission boundaries.

Conclusion

AI is making crypto fraud faster, more tailored, and more scalable, while defenses are still being rebuilt around this new reality. Large exchanges and regulators are deploying their own AI tools and coordination frameworks, but experts warn that the window to get ahead of adversaries is short. For now, the most resilient position is to assume AI enhanced scams are everywhere and design both personal habits and protocol architectures so that no single convincing message or model output can move your assets on its own.