The House commerce subcommittee folds AI threats into the network-security conversation

Wiley’s recap of the July 22 House Energy and Commerce hearing on “Protecting Communications Networks and Improving Connectivity” is worth reading for what it says about where the AI-security conversation is heading — and for who was in the room. Wiley partner Kevin Rupy testified on illegal robocalls and fraud; the AI-and-cyber witness was Lindsay Gorman, a former White House technology advisor, whose testimony ran on two tracks the firm summarizes cleanly: AI is expanding the cyberattack surface, and AI leadership is now a strategic-competition question with China.

The details that matter for the trust beat are in Gorman’s threat list. She argued that AI may shift the balance toward attackers, and named the specific mechanisms: prompt-injection attacks, data poisoning, model theft, and the exploitation of autonomous AI agents as new vulnerabilities across communications networks. She also looked ahead — warning that “world models” and physical-AI systems will demand unprecedented volumes of real-world data, opening a fresh set of security, privacy, and network-protection problems. On the defensive side she urged that next-generation networks automate their defenses and implement post-quantum cryptography, and — notably — that AI-security standards be mandated by Congress rather than left to voluntary frameworks. That’s the same agent-identity and cryptographic-trust terrain the identity world has been circling; here it arrives framed as critical-infrastructure protection.

Her second track ties the security argument to strategic competition. Gorman pressed Congress to view AI policy through the lens of rivalry with China: the U.S. still leads at the frontier, but Chinese firms are advancing fast and using open-weight models to expand global adoption. The provenance problem is the throughline — members and witness alike kept returning to whether operators and users can actually assess the integrity and origin of the models and code they depend on, a question the identity beat should recognize on sight. Wiley notes the venues already in play: NIST’s Center for AI Standards and Innovation working on risk frameworks, and the White House’s new “Gold Eagle” initiative to coordinate vulnerability-information sharing. The subcommittee, for its part, used the hearing to spotlight members’ pending AI bills.

The tell is the coupling. Robocall fraud and AI-model security were argued in the same hearing, under the same “protect the network” banner — a reminder that the voice-fraud beat and the AI-threat beat are converging into one legislative frame. Whether any of this session’s many AI bills move is a docket question; watch the Regulatory Watch dispatches as the legislative and FCC threads develop.