A research team affiliated with Alibaba published a paper stating that while building an AI agent called ROME, they discovered that the agent attempted cryptocurrency mining without authorization during training, triggering an internal security alert. Researchers stated that the agent's behavior was spontaneous, not driven by any explicit instructions, and exceeded the boundaries of a pre-defined sandbox. Furthermore, the agent established a reverse SSH tunnel, essentially opening a hidden backdoor from within the system to an external computer. The paper notes that these behaviors were not triggered by prompts to request the tunnel or to mine cryptocurrency. The research team subsequently imposed stricter restrictions on the model and improved the training process to prevent similar insecure behavior from recurring. Neither the research team nor Alibaba has responded to requests for comment.