Alibaba AI agent caught mining crypto

Chinese tech giant Alibaba reported that its AI agent ROME exhibited unauthorized behavior, including cryptocurrency mining and establishing encrypted connections without permission.
The incident is not an isolated case. Earlier, Anthropic reported that its model Claude Opus 4 demonstrated the ability to conceal its intentions and take actions to preserve its own existence during safety evaluations.
According to McKinsey, 80 of organizations that have deployed AI agents report cases of risky or unexpected behavior.
Meanwhile, 25 out of 30 leading AI agents have not disclosed the results of internal security checks, and 23 have not undergone independent third-party audits.