Two Years After CrowdStrike, We Are Still Afraid Of The Wrong AI Apocalypse – OpEd

Eurasia Review  |  Burak Oktenli

The July 19, 2024 CrowdStrike software update failure crashed approximately 8.5 million Windows machines globally, demonstrating that catastrophic systemic vulnerability stems from automated authority rather than malicious artificial intelligence. This disruption grounded flights, halted emergency rooms, and exposed the dangers of granting unchecked operational permissions to automated systems.

Historically, critical infrastructure failures like the 2015 Ukrainian power grid cyberattack and the 2021 Colonial Pipeline ransomware incident were mitigated only because human operators retained the capability to transition to manual overrides. This reliance on human judgment mirrors historical military precedents, such as the 1983 Soviet early-warning false alarm where human intervention prevented accidental nuclear launch. However, the rapid automation of industrial control systems is steadily eroding this essential manual fallback competence across global networks. To prevent future systemic collapses, organizations must implement strict human-approval gates, audit automated permissions, and actively maintain manual operational protocols as critical national security infrastructure.

Comment

The 2019 cyber intrusion at India's Kudankulam Nuclear Power Plant demonstrated that air-gapped systems remain vulnerable to lateral credential propagation. While the Dtrack malware only compromised administrative networks, it exposed the danger of relying solely on physical isolation. India's National Critical Information Infrastructure Protection Centre must mandate offline operational drills to preserve analogue competency across the electrical grid. Analogue fail-safes must remain entirely independent of digital networks.

To read in details click here