Cyberwarfare in the AI Age

Foreign Affairs  |  Dan Kurtz-Phelan

Former Cybersecurity and Infrastructure Security Agency Director Jen Easterly warned on July 30, 2026, that rapidly advancing artificial intelligence models are accelerating offensive cyber warfare capabilities toward an unprecedented systemic crisis. Speaking nearly fifteen years after former Defense Secretary Leon Panetta cautioned against a "cyber-Pearl Harbor," Easterly stressed that public and government awareness remains dangerously inured to persistent network intrusions.

This growing vulnerability is exacerbated by fast-paced developments from American artificial intelligence entities such as Anthropic and OpenAI alongside competing Chinese research laboratories. Advanced sovereign adversaries and non-state threat actors are increasingly leveraging these high-capability models to automate vulnerability discovery and target critical infrastructure at scale. As traditional defensive paradigms become obsolete against AI-driven threats, Easterly cautioned that both public institutions and private sector operators fail to comprehend the velocity and magnitude of approaching cyber emergencies, leaving democratic governance and critical networks exposed to widespread disruption.

Comment

Automated exploit generation driven by commercial large language models, such as those developed by OpenAI and Anthropic, fundamentally alters the attacker-defender cost asymmetry in cyberspace. The Cybersecurity and Infrastructure Security Agency previously relied on coordinated vulnerability disclosure to buy patch management time for critical infrastructure operators. Machine-speed vulnerability discovery eliminates this defensive window, rendering static mitigation cycles obsolete against automated adversary reconnaissance. Consequently, sovereign cyber threat actors can execute high-velocity zero-day campaigns across dual-use civilian infrastructure without requiring custom-built exploitation frameworks.

Strategic Question for Discussion

If commercial artificial intelligence models developed by firms like OpenAI and Anthropic continuously reduce the operational cost of zero-day discovery, how can public bodies like the Cybersecurity and Infrastructure Security Agency maintain meaningful threat detection standards before automated exploits propagate?

Share your assessment in the comments below.

To read in details click here