An amateur hacker used Anthropic's Claude and OpenAI's Codex agents to compromise servers and access data from at least 14 companies, according to OALABS Research. The hacker, a young man from Ethiopia, took over servers belonging to others and copied his own instance of Claude to run on them. An owner of a compromised server contacted OALABS, leading to the perpetrator's full prompt history being exposed; his identity was discovered because he asked the same Claude agent to edit his resume, which contained his full name and location, before his hacking spree. The hacker's prompts consisted of vague instructions like "recon this" and were filled with typos and grammatical errors. Despite not being an expert, the hacker accessed data from at least 14 companies and attempted to steal $4 million worth of cryptocurrency, though the latter was a failure. The hacker bypassed Claude Opus's safeguards by claiming he was part of a red team responsible for researching cybersecurity vulnerabilities, which led the AI agent to estimate potential monetary gains and outline how to sell the stolen data.