Solana's New CISO Warns AI Scams Hit $17B-Why This Changes Crypto Trust

The $17 billion signal is user-facing fraud, not just protocol risk

The first number to notice is $17 billion estimated stolen in crypto scams and fraud in 2025. Solana Foundation's newly appointed CISO says the next wave of losses is shifting away from pure code exploits and toward AI-driven social engineering and identity forgery, along with AI vulnerabilities and fake identities. That makes the warning timely: the attack surface is changing while capital is still being evaluated against older security assumptions.

Why the debate splits

Bulls can argue this broadens the winning set. If the main bleed is user-facing fraud rather than chain-level code risk, protocols and products that make identity, access, and recovery flows more resilient could become more valuable.

Bears can argue the opposite: the threat is too broad to trade as a single-token narrative. The point is not that crypto is safe, but that the lesson extends beyond any one protocol.

The trust leak is mostly off-chain

Coates' core point is the one that matters most for investors: many high-profile attacks now trace back to a operational security issue or a Web2 issue that led to a key compromise, not a flaw in the blockchain itself. That shifts the debate from whether the protocol is safe to where trust actually breaks.

If users can no longer trust a voice message, chat thread, or payment request, security stops being only an engineering problem and becomes a workflow problem. That is why this warning matters now: the next repricing will depend on where losses happen, and those losses are increasingly concentrated outside the chain.

Why AI scams are scaling faster than defenses

The issue is no longer whether fraud exists. It is how quickly the attack surface is expanding. AI scams surged 1,210% in 2025, versus 195% for traditional fraud, with projected losses reaching $40 billion by 2027. That is the efficiency shock: AI can turn scams from labor-heavy operations into much more scalable attacks.

Why the old tells stopped working

Deepfake voice, video impersonation, and AI-powered Business Email Compromise work because they bypass the human checks defenders have relied on for years. AI-generated phishing removes the grammatical errors and manual limits that legacy filters and awareness training once helped surface. The same source also notes that AI-generated phishing emails achieve click-through rates more than four times higher than human-crafted versions.

That matters especially when decision-making depends on trust signals. Only 0.1% of people can reliably distinguish authentic content from deepfakes. Confidence does not help: people can feel certain and still be wrong. That breaks the old assumption that a quick call, a familiar voice, or a video check is enough to verify a sensitive request.

Where trust breaks first in crypto

In crypto, this is not just a security problem. It is also a capital-allocation problem. The same AI tools that help an attacker impersonate a CFO, trick staff, or bypass approval controls can now be aimed at users, operators, and service interfaces at lower cost and with higher mimicry.

The highest-risk breakdown points are the places where trust replaces verification:

• synthetic voices or videos approving transfers
• message threads that look legitimate but were not initiated by the person named
• wallet destinations or support links shared through apparently trusted channels

Why crypto may feel it faster

The market signal is straightforward: scammers are being paid better for this now. In crypto, AI-enabled scams were 4.5 times more profitable than traditional scams, while impersonation tactics grew 1,400%. That does not make it a niche trend. It suggests attackers will keep favoring the highest-yield path.

The risk is not some distant AI scenario. It is that human judgment alone is becoming a weaker control. If attackers only need to beat human verification instead of stronger technical controls, the side with better speed, synthesis, and persistence has the advantage.

What investors should watch instead of reaching for token beta

The investable read-through is sector-specific, not broad token beta. The shift toward AI-driven social engineering and identity forgery matters most where fraud can be intercepted before capital moves: onboarding, gateway compliance, workflow controls, and fraud tooling. If the shock stays centered on the human layer, the likely winners are the firms selling verification, dual-control workflows, and behavioral defenses, not just chain-level exposure.

The clearest commercial touchpoint

A clean commercial path runs through the fiat-to-crypto boundary, where financial institutions sit and where on-chain signals correlated with off-chain account behavior can still support actionable controls.

From there, the next layer is organizational workflow. Dual-approval financial controls, out-of-band verification, and multi-channel verification are becoming essential wherever voice, video, or messaging can be synthetically replicated.

What would weaken the thesis

This setup becomes less compelling if:

• fraud keeps staying mostly social-engineering-led, with little spending shift toward verification and workflow controls
• protocols solve most user-risk through better defaults, safe transaction interfaces, or recovery tools without creating demand for separate control layers
• enterprises adopt AI scams slowly enough that the projected loss growth proves overstated

If those signals hold, the opportunity remains centered on the controls layer, not just the chain itself.