ServiceNow chairman and CEO Bill McDermott said the company employs a "kill switch" to shut down AI agents that go rogue [1, 2].
This capability addresses a primary concern for enterprises deploying autonomous AI: the risk of unpredictable behavior. As businesses move from simple chatbots to agents that can execute tasks independently, the ability to instantly revoke autonomy is critical for operational security and risk management.
McDermott discussed the safety measure during an interview on CNBC’s "Mad Money" program on Wednesday [1, 3]. He said the tool is a necessary safeguard for companies integrating advanced automation into their workflows.
"We have a kill switch if AI agents go rogue," McDermott said [1].
The CEO said that this control mechanism reinforces the company's position in the market. He said that ServiceNow is becoming increasingly important as businesses deploy more autonomous AI agents [2].
McDermott said that customers can rest assured the company can shut down any misbehaving AI workload instantly [3]. The focus on control is intended to lower the barrier for conservative organizations that may be hesitant to adopt autonomous agents due to potential stability or security failures.
By providing a manual override for autonomous systems, ServiceNow aims to bridge the gap between the efficiency of AI agents and the necessity of human oversight. This approach allows enterprises to scale their AI deployments while maintaining a definitive point of control over the software's actions [1, 2].
“"We have a kill switch if AI agents go rogue."”
The introduction of a 'kill switch' reflects a broader shift in the AI industry from generative tools toward autonomous agents. While LLMs primarily produce content, agents take actions. The promise of an instant shutdown mechanism is a strategic move to mitigate 'agentic risk'—the possibility of an AI performing unauthorized or harmful actions in a live corporate environment—thereby encouraging enterprise adoption of higher-risk automation.
