Industry leaders are warning that AI agents could push crypto exploit losses far beyond today's benchmarks, potentially making the sector's current billion-dollar hacks look like “pennies” as automated attackers lower the cost and raise the scale of on-chain theft.

The warning, reported by The Block, frames AI-driven crypto hacks as a forward-looking risk rather than a single completed exploit. At issue is the possibility that autonomous and semi-autonomous agents could change the economics of attacking crypto protocols, at a time when both the defensive and offensive sides of the industry are already adopting more automation.

Why industry leaders think AI agents could multiply hack losses

The core argument is about automation economics. When an agent can handle work that previously required skilled human operators, the marginal cost of attempting an attack falls — and the number of attempts a single actor can run rises accordingly.

Speed compounds that shift. If agents can execute an exploit faster than defenders and monitoring systems can respond, more value can be drained before a protocol pauses contracts or freezes funds. That gap between attack speed and response speed is why the threat is framed as a step-change rather than an incremental one.

Notably, the warning is attributed to multiple sector voices rather than a single isolated opinion. That breadth is what gives the “pennies” comparison its weight as a risk signal for crypto security specifically — not as general AI hype.

How AI agents could change the attacker playbook

Lower manual workload means attackers can test more targets and more strategies in parallel. That widens the pool of people capable of running a credible attack, since expertise that once gated the process can now be partly delegated to software.

Agentic systems could also compress the phases of an attack — from reconnaissance and code review to phishing and exploit iteration — into faster loops. The risk spans two distinct surfaces: direct smart-contract exploits and adjacent social-engineering attacks that target users rather than code.

The same automation pressures already shaping legitimate crypto tooling, visible at industry events focused on the intersection of Web3, digital assets and AI, cut both ways — giving attackers the same efficiency gains that builders are pursuing.

What protocols, investors, and users should watch

For protocols, the practical priorities are monitoring, audit depth, and response speed. If agent-driven attacks execute faster than traditional review cycles, the value of real-time detection and rapid incident response rises relative to slower, periodic audits.

For users, the exposure is more familiar: wallet hygiene, token approvals, and phishing awareness. Faster, cheaper social-engineering campaigns raise the odds that a given user will encounter a convincing attempt.

The capital flowing into the sector — including the pace of crypto venture funding in the first half of 2026 — underscores how much value now sits behind the smart contracts these warnings target. AI-linked assets such as Render illustrate how tightly the AI and crypto narratives have converged.

The evidence available describes rising risk, not a guaranteed outcome. The message from industry leaders is one of preparedness before agent-driven attacks mature further — not a prediction that a specific exploit is imminent.