AI agents find $4.6M in blockchain smart contract exploits

kept by eddie dec 02, 25

AI agents autonomously exploited smart contracts in simulation, generating $4.6 million in simulated stolen funds, demonstrating a concrete lower bound on economic harm from AI-driven cyberattacks.

An Unlikely Source of Crypto Innovation: Afghanistan - The New York Times
An unlikely source of crypto innovation, Afghanistan, sees a start-up building blockchain tools to transform humanitarian aid despite Taliban suspicion of the internet.
also on #blockchain
Notion AI: Unpatched data exfiltration
Notion AI allowed data exfiltration via indirect prompt injection in March 2025, where edits were saved before user approval, enabling attackers to steal sensitive document contents.
also on #ai-security
BrowseSafe: Understanding and Preventing Prompt Injection Within AI Browser Agents
Perplexity AI introduces BrowseSafe, a defense architecture, benchmark, and detection model for securing browser agents from prompt injection attacks.
also on #ai-security
Without Benchmarking LLMs, You're Likely Overpaying 5-10x
Benchmarking LLMs on actual task-specific prompts can find cheaper alternatives that match performance, saving 5-10x on API costs.
also on Claude Opus 4.5, GPT-5
Claude Codes #3 - by Zvi Mowshowitz
Claude Code and Claude Cowork dominate AI coding tools, with Anthropic's safety approach enabling permissive agents, but users warned against over-optimizing setups.
also on Claude Opus 4.5, Anthropic