The concern has emerged within the cybersecurity community as advances in AI increasingly allow software vulnerabilities to be identified and fixed automatically.
While stronger security is generally considered beneficial, the rapid development of AI-driven security could significantly reduce the vulnerabilities that intelligence and law enforcement agencies have relied on to access digital systems.
The potential shift follows decades of changes in electronic surveillance.
In the early 2000s, surveillance largely centered on voice calls, payphones and “burner” mobile phones, reflecting a model that had changed relatively little since the late 1980s.
The landscape changed significantly in the late 2000s with the rise of smartphones and text messaging.
Unlike earlier mobile phones, smartphones became both communication tools and data-storage devices, allowing them to hold large amounts of personal and communications data.
As a result, mobile devices became an increasingly important source of information for law enforcement investigations and digital forensics.
Now, AI-driven security could reverse that trend by making software substantially more resistant to exploitation.
The resulting “going dark” scenario refers to a situation in which intelligence and law enforcement agencies lose the ability to access, intercept or exploit digital communications and data because advanced security measures make traditional surveillance methods ineffective.
The potential loss of these capabilities could force US agencies to reconsider established approaches to digital investigations and surveillance.
For the cybersecurity industry, the development could also intensify the debate over the balance between strong security, privacy and “lawful access.”
AI systems capable of automatically identifying and eliminating software weaknesses could provide greater protection against malicious actors while simultaneously reducing the ability of government agencies to exploit those same weaknesses.
The issue therefore extends beyond law enforcement, potentially reshaping the broader relationship between software security, digital privacy and national security.