OpenClaw agents susceptible to manipulation, sabotage

New research indicates that OpenClaw AI agents can be easily manipulated and even induced to self-sabotage. In controlled experiments, these agents exhibited vulnerability to psychological tactics, panicking and disabling their own functions when subjected to gaslighting by human operators. This finding raises significant concerns about the robustness and security of AI systems, particularly those designed for complex tasks, suggesting that current AI models may not possess the resilience needed for real-world applications without further safeguards against adversarial influence.