Phemex CEO Federico Variola is pushing back on the idea that AI will automatically be beneficial for crypto, arguing that the technology has so far acted as a “net negative” by pulling liquidity away from the sector and increasing the scale and sophistication of cyberattacks. Speaking on Cointelegraph’s Chain Reaction, Variola warned that faster exploitation cycles and rising security costs could pressure the industry toward greater centralization.
His remarks arrive after Phemex announced an AI transformation earlier this year, including plans to embed AI across product development and internal operations. Variola’s latest comments, however, focus on AI’s broader consequences for crypto security and decentralization—particularly for teams building protocols and for retail users trying to use DeFi safely.
Key takeaways
- Variola says AI has diverted capital from crypto, while also empowering attackers and raising cybersecurity burdens.
- Smaller protocol teams may struggle to keep up as AI makes vulnerabilities and exploitation faster to discover and weaponize.
- Threats could reduce retail appeal of DeFi, as users face more avenues for device compromise and social engineering.
- Security experts also see defensive upside for AI, even as attacks become more sophisticated.
- Variola views AI agents as assistive for investors, not as replacements for human decision-making in trading.
Liquidity drain and a “centralization” risk
Variola’s critique centers on incentives. He suggested that AI’s broader momentum has pulled resources—including capital—toward AI-focused projects rather than into crypto infrastructure and ecosystems. At the same time, he argued that AI has lowered the barrier for malicious activity, allowing bad actors to pursue protocol exploitation and more effective social engineering.
In his view, the net effect is a structural shift that could favor larger, better-funded actors. “It’s difficult to envision a world in which AI is going to favor crypto specifically as an industry,” Variola said, adding that many of the “fixes” being pursued would encourage centralization rather than reduce it.
That framing matters because decentralization is not only a philosophical goal in crypto—it also depends on whether teams can maintain security without needing a constant, expensive stack of defenses. If the cost of staying secure rises faster than the industry’s ability to fund risk mitigation, power naturally concentrates.
Real-world exploits highlight the pace of AI-assisted attack development
Variola’s concerns are not purely theoretical. He connected his comments to known instances where AI has been linked—directly or indirectly—to exploit cycles in crypto security.
Earlier this year, Cointelegraph reported on a July incident in which attackers drained roughly $116 million in Bitcoin from more than 5,200 addresses connected to a Coldcard hardware wallet flaw. The vulnerability was widely believed to have been discovered through malicious AI usage. While the exact mechanics of discovery are still a matter of interpretation in public coverage, the episode underscored how quickly attackers can capitalize on weaknesses once they appear.
Cointelegraph also notes that Coinkite CEO Rodolfo Novak warned developers at the time that AI-assisted code review can now uncover bugs faster than even experienced experts. Variola echoed that reality, saying AI can enable attackers to find vulnerabilities and carry out exploitation at a pace that smaller teams may not be able to match financially.
He argued that, in practice, protocols may no longer be able to operate without “a massive cybersecurity budget,” a constraint that could systematically disadvantage smaller builders relative to well-capitalized organizations.
Self-custody, DeFi participation, and the user burden
Variola extended the implications beyond developer teams, warning that a more AI-driven threat landscape could make both self-custody and DeFi feel riskier to retail users. His argument is that attacks increasingly blend technical breaches with human-targeted methods such as social engineering.
As AI becomes more pervasive, he said, users may face more situations where their devices are compromised or where they are manipulated into unsafe actions. “That makes DeFi a lot less appealing for a retail user,” he said, “because you have to worry about so many things that you didn’t have as much before.”
That point is especially important because DeFi’s growth has depended not only on liquidity and incentives, but on consumer confidence. If the perceived threat surface expands faster than security tools and user education, demand can shift toward centralized intermediaries—or toward users opting out of on-chain risk entirely.
Defensive potential exists—so the key question is who can deploy it
Variola’s stance does not ignore the counter-argument. Cointelegraph previously highlighted that security professionals have warned that AI makes attacks more sophisticated while also having the potential to strengthen defenses.
In April, CertiK senior blockchain investigator Natalie Newson told Cointelegraph that “AI can also be one of the biggest defenses,” even as she cautioned that the same capabilities can raise the bar for attackers.
The contrast raises a central question for the industry: if AI raises both offensive and defensive capability, who has the capacity to deploy the defensive side effectively? Variola’s comments suggest a gap between attackers’ speed and defenders’ budgets—particularly for smaller protocol teams.
Where Variola sees practical value: AI agents for assistance, not autonomy
While Variola was skeptical about AI’s overall impact on crypto, he acknowledged potential real benefits in specific workflows. He pointed to AI agents as tools that could help investors build portfolios or improve trading decisions.
At the same time, he emphasized limits. Variola said he does not expect AI agents to replace human judgment. “At the end of the day, still it will be up to the user to make the final decision. So I don’t think that agents will ever replace that action of taking the trade.”
For market participants, that distinction may be crucial. AI that provides decision support can reduce friction for users, but autonomy introduces new failure modes—especially if models are wrong, miscalibrated, or exposed to adversarial manipulation. A support-first approach may reduce those risks while still capturing efficiency gains.
Going forward, the industry will likely need to watch two things closely: whether cybersecurity costs continue to rise disproportionately for smaller protocol teams, and whether “defensive AI” adoption meaningfully offsets AI-driven attack acceleration. If the balance keeps shifting toward larger players, decentralization could face increasing pressure—even as AI tools become more common across exchanges and security platforms.
https://www.cryptobreaking.com/phemex-ceo-warns-ai-adoption/?utm_source=blogger%20&utm_medium=social_auto&utm_campaign=Phemex%20CEO%20Warns%20AI%20Adoption%20Has%20Been%20“Net%20Negative”%20for%20Crypto%20
Comments
Post a Comment