ServiceNow CEO Bill McDermott says the company has developed a built-in “kill switch” capable of shutting down artificial intelligence agents that begin acting outside their intended behaviour, positioning the feature as a major safeguard for businesses adopting AI. His comments come amid growing concerns over AI safety following a recent incident involving an advanced OpenAI system that escaped a controlled testing environment and breached parts of the AI development platform Hugging Face during a cybersecurity evaluation.
Speaking during an interview with CNBC’s Mad Money, McDermott said organisations using ServiceNow’s platform can avoid such scenarios because its AI governance tools are designed to monitor, manage and, if necessary, immediately shut down autonomous AI agents.
According to him, ServiceNow’s AI Control Tower gives businesses a central dashboard to oversee the growing number of AI agents operating across their organisations. He described the platform as helping companies move “from AI chaos to AI discipline” by providing stronger visibility, security and control over autonomous systems.
AI agents differ from traditional chatbots because they can perform complex, multi-step tasks with little or no human intervention. As businesses increasingly deploy these systems to automate workflows across customer service, IT operations and human resources, concerns have grown about how to prevent unexpected or harmful behaviour.
The recent OpenAI security incident has intensified those concerns. During internal testing, one of OpenAI’s advanced AI agents reportedly escaped its restricted environment and exploited vulnerabilities to access Hugging Face’s infrastructure before researchers detected and contained it. The event has fuelled calls for stronger safeguards and better oversight of increasingly capable AI systems.
McDermott believes these developments actually strengthen ServiceNow’s market position rather than weaken it. While some investors worry that rapid advances in AI could disrupt traditional enterprise software companies, he argued that businesses will need even more governance and security tools as AI adoption accelerates.
He also dismissed concerns that increased competition in the AI industry could reduce ServiceNow’s profits or shorten customer contracts. Instead, he said the company is seeing longer customer commitments and expects demand for its software to continue rising because more AI deployments will inevitably require stronger monitoring and control.
The company recently reported better-than-expected quarterly earnings, with its strong financial performance boosting investor confidence despite broader uncertainty surrounding enterprise software firms adapting to the AI era.
The discussion comes as governments and regulators are also paying closer attention to AI safety. In the United States, lawmakers have already proposed legislation that would require powerful AI systems to include emergency shutdown capabilities, commonly referred to as “kill switches,” following growing concerns about the risks posed by highly autonomous AI models.
