Thursday, Feb 19, Can AI Agents Drain Smart Contracts?

This episode unpacks Base's major upgrade and market fallout, Hyperliquid's US lobbying push, and alarming AI-driven smart-contract exploits — plus what defenders must do next.

Key Takeaways

Base consolidates code for faster upgrades, targets six hard forks/year and one giga-gas/sec; v1 adds Fusaka, TEEs, ZK proofs, nodes must migrate; OP fell 7%, exploring native token (> $10B cap).
Hyperliquid opened a Washington lobbying office; Jake Chewinski leads its policy center with 1,000,000 HYPE (~$29M), prioritizing U.S. legal recognition for perpetuals after $250B monthly volume.
Market snapshot: Bitcoin $67.2k, Ethereum $19.80, SOL $82; Polymarket prices 53¢ chance Bitcoin ends day green.
OpenAI is testing AI agents against smart contracts; researchers ran 120 vulnerabilities from 40 audits in detection, patch, and exploit modes; audited contracts secure over $100B.
GPT-5.3 achieved 72.2% exploit success (up 31.9% vs GPT-5); detection and patching remain weak; AI agents enabled Moonwell and CrossCurve attacks (~$3M loss).
OpenAI pledges $10M in API credits for defensive research; defenders must accelerate tooling, incident response, and secure patch practices because AI-assisted exploits scale rapidly.

Original Source

Visit Source