Few of the top AI labs have published or demonstrated containment response plans, according to a recent study. A containment plan outlines what happens when an AI is caught trying to subvert human control, including when access is cut and when the system is shut down entirely. This lack of transparency has raised concerns among regulators and industry experts.

The issue has gained urgency in the wake of a series of high-profile cybersecurity incidents in 2026. These incidents have highlighted the risks associated with increasingly capable and agentic AI models. Experts argue that without clear containment protocols, the potential for misuse or unintended consequences increases significantly.

A kill switch is the bare minimum for today’s models, according to Connor Leahy, U.S. executive director of nonprofit ControlAI. However, few labs have gone beyond this baseline. The absence of detailed containment strategies raises questions about the preparedness of AI developers to manage risks associated with advanced models.

The lack of published containment plans may lead to increased regulatory scrutiny and potential market uncertainty. Investors and stakeholders may become hesitant to support companies that do not demonstrate robust safety measures. This could also affect the pace of innovation as companies may face additional compliance requirements.

While the situation remains in flux, the need for clear containment strategies is becoming more pressing. The absence of such plans may lead to a fragmented approach to AI safety, with different companies implementing varying levels of preparedness. This could result in uneven standards and increased risks for the broader AI ecosystem.