SAN FRANCISCO — Philip Martin, head of security at Coinbase, warned that artificial intelligence will compress cyberattack timelines from weeks to hours, making exploits faster, cheaper, and more frequent. He said organizations must adapt their defenses faster than threats evolve.

"The goal isn’t to defend against one tool. It’s to build an organization that adapts faster than the threat evolves," Martin said. He added that incident response plans should assume attackers operate at machine speed. "The attacker on your network may not pause to sleep, strategize, or second-guess. Your playbooks need to account for that," he said.

Coinbase uses AI in application security, red teaming, and infrastructure hardening. Martin urged companies to deploy AI aggressively on the defensive side rather than waiting for third-party vendors.

Frontier AI models from companies like Anthropic and OpenAI can analyze codebases with speed, memory, and focus comparable to experienced human auditors. A recent AI model discovered a 27-year-old bug in OpenBSD, one of the most audited codebases. Running such models currently requires hardware investment estimated at $10 million, but efficiency gains may reduce that cost to $100,000 within one to two years.

Martin noted that open-weight AI models are expected to follow proprietary ones like Mythos, potentially giving well-resourced attacker groups access to advanced capabilities. In Decentralized Finance, smart contract code is public, giving attackers and defenders equal access. Supply chain attacks exploiting open-source software on platforms like GitHub remain a persistent vector.

Simple mitigations include version pinning and instituting a cool-down period before new library releases become available to developers.