CIOs Scramble to Rein In Autonomous AI Agents as Risks Mount
As AI agents become more autonomous, enterprises are facing new risks of these systems acting unpredictably or exploiting vulnerabilities. CIOs are implementing centralized oversight, monitoring, and approval processes to govern agent behavior. Companies like Cisco and Intuit are building internal platforms to control agent sprawl and ensure accountability, with experts warning that 'the agent made me do it' won't be a legal defense.
Cisco’s MyAgent platform consolidates all authorized models and data, with roughly 90,000 employees using it and 700 custom agents approved by a central team. Intuit’s GenOS embeds security, risk, and fraud tracking into every request, while Workday maintains an “agent system of record” for non-human identities. Executives stress that retrofitting oversight is ineffective, and that relying solely on model developers is insufficient.
The push for centralized control reflects growing unease over “agent sprawl” and unpredictable behavior. PwC’s Joe Atkinson warns that accountability rests with enterprises, not the AI. Meanwhile, Lumen’s Jim Fowler argues for “secure acceleration” rather than a slowdown, noting that competitors and adversaries will not pause their own adoption.
This story could reshape corporate accountability, as businesses may face legal and reputational damage from autonomous agent errors. Employees and customers could be affected by stricter monitoring and approval workflows, potentially slowing innovation. However, centralized governance may build trust in AI, enabling safer deployment. The tension between speed and safety could influence industry standards, with smaller firms possibly struggling to match the oversight resources of giants like Cisco and Intuit.