AI-Driven Hacks Surge, DeFi Deemed 'Unsafe'
Anthropic reports 67% of banned accounts used AI for cyberattack prep, fueling a surge in crypto theft to $629.7M in April. OpenZeppelin founder calls all of DeFi unsafe as AI models gain autonomous hacking capabilities, with Anthropic's upcoming Mythos model finding 10,000 vulnerabilities.
Quick Take
67% of Anthropic-banned accounts used AI for cyberattack preparation.
Crypto theft surged to $629.7M in April, linked to AI-assisted hacks.
AI now assists with post-compromise lateral movement in 6.5% of cases.
Upcoming Mythos model found 10,000 vulnerabilities, heightening crypto risks.
Market Impact Analysis
BearishAI-enabled cyberattacks lower the barrier to sophisticated hacks, increasing frequency and severity of crypto thefts, which erodes market confidence and may prompt regulatory scrutiny.
Speculation Analysis
Key Takeaways
- 67% of Anthropic's banned accounts were leveraging AI to plan and execute cyberattacks.
- April crypto theft hit $629.7M — the highest since February 2025 — with AI likely a key enabler.
- 6.5% of accounts used AI for lateral movement, a technique once reserved for skilled hackers.
- Anthropic's upcoming Mythos model uncovered 10,000 vulnerabilities, signaling escalating DeFi risks.
What Happened
AI firm Anthropic revealed that over two-thirds of accounts it banned for policy violations in the past year were using its models to prepare for cyberattacks. Out of 832 examined accounts, 560 leveraged AI to write malware, identify vulnerabilities, and orchestrate attacks. This data coincides with a brutal April for crypto security: hacks stole $629.7 million, the highest monthly total since February 2025. OpenZeppelin founder Manuel Aráoz responded by declaring "all of DeFi unsafe," citing AI models' growing ability to detect smart contract flaws. The findings signal a dangerous shift where AI is becoming a core tool for cybercriminals targeting digital assets.
The Numbers
Anthropic's investigation of 832 banned accounts found 560 using AI for attack preparation—a striking 67%. In 6.5% of cases, AI assisted with lateral movement, a post-compromise technique historically limited to advanced threat actors. The risk escalated over time: accounts classified as medium risk or higher jumped from 33% in the first six months to 56% in the latter half. Meanwhile, crypto theft in April reached $629.7 million, with analysts attributing part of the spike to AI-assisted exploits. These figures underscore how AI is amplifying both the frequency and sophistication of attacks.
Why It Happened
AI models have drastically lowered the barrier to entry for cybercriminals. Tasks like vulnerability scanning, malware generation, and credential theft—once requiring deep technical expertise—can now be automated. Anthropic observed that there is "little correlation between the skill of a threat actor and how many techniques they use." Even state-sponsored groups are deploying autonomous AI agents that exploit, steal, and make decisions with minimal human input. The rise of open-source AI models and prompt engineering breakthroughs has further democratized these capabilities, turning DeFi protocols into high-value targets.
Broader Impact
The surge in AI-powered hacks threatens to erode trust in decentralized finance and attract harsher regulatory scrutiny. If DeFi is widely perceived as unsafe, institutional adoption could stall. Security firms may face pressure to integrate AI-defense tools, while regulators might fast-track stricter cybersecurity requirements. Anthropic's warning that autonomous attacks will increase as AI agents become more capable suggests the industry is entering a new era where the attack surface expands faster than defenses can keep up.
What to Watch Next
- Mythos model release: Anthropic's new model already exposed 10,000 vulnerabilities; its public debut could lead to a wave of exploits unless patching accelerates.
- Regulatory response: Watch for statements from the SEC or CFTC addressing AI-enabled cyber threats in crypto markets.
- DeFi security upgrades: Protocols may rush to implement real-time AI monitoring and bug bounty programs to counter the threat.
This article is for informational purposes only and does not constitute financial advice.
Always late to trends?
Join for the latest news, insights & more.
Disclaimer: Bytewit is an independent media outlet that delivers news, research, and data.
© 2026 Bytewit. All Rights Reserved. This article is for informational purposes only.
