AI Attack Speeds Far Outpace Human Defenses: How U.S. Power Grids, Banks, and Hospitals Maintain the Security Bottom Line

Author: Anne Neuberger, a16z

Translated by: TechFlow

TechFlow Summary: AI is drastically shortening the window from when a software vulnerability is discovered to when it is exploited. Meanwhile, in the US critical infrastructure sectors such as power, banking payments, water supply, telecommunications, and hospitals, a vast number of operational technology (OT) systems running physical equipment have been in service for over 20 years and cannot possibly be patched at this speed.

Leveraging her experience as former US Deputy National Security Advisor for Cyber and Emerging Technologies, a16z General Partner and Head of Global Affairs Anne Neuberger proposes a public-private cyber defense framework: setting national targets via the White House Super Intelligence Force, having industry-leading CEOs sign commitments, and repurposing existing Industry Sector Information Sharing and Analysis Centers (ISACs) to scale AI-driven defenses.

The core logic of this plan is not to completely replace aging equipment, but to first identify hundreds of core operators and shared suppliers whose failure would trigger nationwide cascading outages, and contain risks within six months through network isolation, 72-hour manual run drills, and AI-based compensatory controls.

Today, a16z General Partner and Head of Global Affairs Anne Neuberger wrote an opinion piece in The Wall Street Journal stating:

AI is continuously compressing the time between vulnerability discovery and exploitation. Defenders still responding at human speed will inevitably lose this race.

The same AI can be used by defenders: discovering exposed systems, assessing which targets are most critical, and generating remediation plans. For aging devices that cannot be patched, AI can generate alternative mitigation measures, such as firewall rules or parameter configuration adjustments. The question is whether we can deploy them quickly enough.

Luckily, the most significant national-level risks are highly concentrated. Risks that could trigger nationwide cascading failures mainly focus on a few hundred operators and their shared suppliers: major transmission substations, Fedwire and ACH payment systems, water facilities serving large cities, core telecommunications networks and 911 systems, regional trauma centers, and electronic health records vendors.

Smaller operators are equally important, often serving as weak points in defense, but prioritizing the protection of this critical minority can buy valuable time to subsequently patch the rest of the systems.

Drawing on her tenure as the US Deputy National Security Advisor for Cyber and Emerging Technologies, Anne outlines a new model for advancing national cyber resilience: reporting to the White House Super Intelligence Force, and led by CEOs of core US companies responsible for operating critical infrastructure.

Executive Summary

The US must confront the significant risk that cyberattacks could paralyze the critical infrastructure upon which millions rely. This cyber defense plan clarifies the required governance architecture and specific actions, aiming to identify and control the risk exposure of these hundreds of core operators and shared suppliers, preventing their failures from triggering nationwide cascading effects.

This action plan aims to: (a) establish robust cybersecurity protections across industries while accounting for numerous legacy and unmanaged systems; (b) leverage Super Intelligence (SI) technologies to discover and patch vulnerabilities in industry-specific, widely deployed code; (c) build AI guidelines or protective frameworks for unresolved corporate vulnerabilities to quickly identify required patches when new models are released.

The second phase of this effort will build digital replicas ("digital twins") of critical shared systems across industries, enabling continuous testing of shared systems.

This public-private partnership fully mobilizes leading enterprises across industries and clarifies their responsibilities, as they best understand their sector's most critical operational risks. Without their participation, given the massive scale of US infrastructure, rapid and widespread progress would be nearly impossible.

The White House Super Intelligence Force will establish a CEO Committee responsible for overall outcomes. ISACs across industries will transform into AI-driven execution engines, delivering results to sector-designated regulators.

Scope: Five key sectors: power, banking & payments, water, telecom, and healthcare/hospitals.

Core Targets: Hundreds of operators and shared suppliers whose failures would trigger nationwide cascading outages.

Technical Core: Use AI to discover vulnerabilities and generate tested alternative protective measures at machine speed, eliminate internet exposure for control systems, implement network segmentation, and validate 72-hour manual operational capabilities.

Simultaneously, define the national-scale defensive capabilities that telecom operators, Infrastructure-as-a-Service (IaaS) providers, and cloud networking vendors (e.g., Cloudflare) can offer, along with the legal protections required to activate these national cyber defenses.

Governance Core: CEOs sign off on sector targets and report progress monthly via ISAC scorecards. Large operators assist smaller ones.

Potential Legal Revisions Needed: Liability immunity protections for information sharing, and potential "safe harbor" mechanisms for participating companies achieving outcome-oriented targets.

Threats and Regulatory Gaps

AI is continuously compressing the timeframe from vulnerability discovery to exploitation. Meanwhile, a vast amount of current critical infrastructure cannot complete patching within this timeframe. Operational Technology (OT), which controls physical equipment, often operates continuously for over 20 years, with some systems facing downtime interruption risks if patched.

Existing regulatory rules were not designed for this threat. They rely on fixed patch windows and compliance-driven mindsets, evaluating whether controls are formally in place rather than actual exposure size and risk containment speed.

Influence-Oriented Organizational Architecture

National Governance Model for Rapidly Scaling Cyber Defense

Reporting to the White House Super Intelligence Force, led by CEOs, and operating through existing Sector Coordinating Councils (SCCs), Information Sharing and Analysis Centers (ISACs), and regulators.

The US currently lacks a governance model capable of driving targeted, rapid action to protect the most critical infrastructure. This plan establishes that model within a public-private framework. In short, the White House Super Intelligence Force sets national targets, CEOs make industry commitments, and SCCs and ISACs drive implementation and measure outcomes within each sector.

The US actually possesses a governance system to drive such work. Established twenty years ago via executive order, the framework mandates a strategic Sector Coordinating Council (SCC) and an operational Information Sharing and Analysis Center (ISAC) for each industry, but this system urgently needs modernization.

Both SCCs and ISACs are privately led non-profit entities operating under federal directives and established public-private partnerships. The most effective operating bodies typically feature a dual CEO and Chief Information Security Officer (CISO) architecture to balance cross-sector strategic decision-making with operational execution.

Currently, ISACs for the power and financial services sectors have achieved C-suite engagement, but those for water, telecommunications, and hospital sectors have not yet established this tier.

The newly established White House Super Intelligence Force should fully leverage these ISACs to coordinate engagements with critical infrastructure sectors. In industries lacking a CEO tier, the SIF needs to partner with the most influential corporate CEOs to upgrade relevant ISACs, establish CEO tiers, and drive necessary actions forward.

Beyond existing ISACs, this plan will also activate two additional forces: (a) cloud providers, cybersecurity vendors, and telecom operators to identify scalable technical defenses under appropriate liability protections; (b) AI labs and tech companies to partner with critical infrastructure firms to scan their code and firmware, generating remediation plans and alternative mitigation measures.

AI labs should also partner with industrial control system manufacturers to build digital twins for standard equipment, allowing remediation testing before actual integration into live grids or water plants; findings will be consolidated into the government's newly established AI-enabled vulnerability exchange hub, Gold Eagle, expanding its scope from software to industrial equipment.

Equipment manufacturers (many of whom are foreign entities) must be required to act on vulnerability disclosures within specified deadlines, making this a prerequisite for selling equipment to critical operators. Success metrics should depend on how quickly vulnerabilities are contained and whether core services remain operational during attacks.

The following chart summarizes the participants and their responsibilities in this national cyber defense plan:

How Accountability Mechanisms Work

Sign Commitments. Each CEO on the committee must sign off on sector targets: zero internet exposure for control systems, passing 72-hour manual operation drills, and containing critical vulnerabilities within designated timeframes (patching or deploying compensatory controls).

Monthly Scorecards. Each ISAC submits anonymized sector-wide performance reports to the White House SIF. Every CEO can view their company's standing within the sector. If a sector falls behind targets, a CEO-level review triggered by the SIF head will be initiated.

Board Certification. For publicly traded companies, the board certifies outcome-oriented progress metrics, aligning with current SEC cybersecurity disclosure rules. Regulators will accept this certification in lieu of redundant audits, incentivizing active corporate participation.

Large Firms Support Small Ones. Major utilities support cooperatives and municipal facilities within their service areas; large hospital systems aid rural clinics; major banks fund community banking tools via FS-ISAC; nationwide telecom operators provide fallback support for rural carriers.

Cross-Sector Joint Exercises. All sectors heavily rely on power and communications. Therefore, within 3 to 6 months, the CEO Committee will organize a joint exercise modeled after GridEx, bringing together power, telecom, banking, and hospital sectors to collectively defend against AI-speed cyberattacks.

Functional Transformation of Sector ISACs

Currently, ISACs primarily share threat alerts at the CISO level. Under this plan, their functions will expand to coordinate and drive cross-industry initiatives:

Share AI Tools: Provide detection rules, exposure scanning for member units, and AI-generated alternative mitigation strategies (finalized and tested on digital twins acting as true software copies of real systems to ensure operational reliability).

Coordinate Remediation Channels: Establish unified vendor liaison channels for AI-discovered vulnerabilities affecting entire sectors.

Scale Mutual Aid Mechanisms: Further expand upon the Power Sector Cyber Mutual Assistance program and the Banking Sector's data recovery standard, Sheltered Harbor.

Metric Aggregation: ISACs centrally generate scorecards, requiring companies to report data only once.

Note: Without legal backing, CEOs will not authorize sensitive data sharing, and companies may face legal liability for self-identified vulnerabilities. Therefore, the Super Intelligence Force must explicitly define legal protections for both scenarios: the former addressed by reauthorizing the Cybersecurity Information Sharing Act of 2015, and the latter implemented through time-bound liability protection mechanisms.

Technical Strategy

Strategic Technical Pathway: Contain Risk, Validate Resilience, Scale via AI

The overarching technical strategy prioritizes protecting the most critical targets, implementing risk containment for unpatchable systems, and leveraging AI to enable a small group of defenders to cover thousands of operators.

Tier Targets by Industry and Identify Core Participants

Rank operators by the severity of outage impact and the number of dependent entities, rather than solely by company size. Shared suppliers often carry greater systemic risk than single operators. Below are recommended implementation pathways for the initial five sectors.

Power: Major transmission substations, balancing authorities, and black start plants used to restart the grid after widespread blackouts.

Banking & Payments: Fedwire, ACH, CHIPS, clearinghouses, and widely shared cloud services and core banking system vendors.

Water: Water agencies serving large populations (50,000+ US residents), especially their chemical dosing control systems.

Telecommunications: Internet routing, core mobile networks, 911 emergency calling systems, and time synchronization signals relied upon by all sectors.

Hospitals: Regional trauma centers, electronic health record (EHR) system vendors, and medical device manufacturers.

Isolate Unpatchable Systems

Audit and Close Exposure Surfaces: Use AI-assisted scanning to discover all internet-facing control systems, VPNs, and remote access paths. Isolate them one by one, place them behind monitored access points, or protect them with phishing-resistant multi-factor authentication.

Unidirectional Communication Links: Deploy hardware devices that only allow data to flow outward from the most critical control networks.

Reduce Blast Radius: Separate IT networks from control networks and segment them by plant or site to prevent single-point breaches from spreading across an entire utility or healthcare system.

Validate Manual Operations: Each core target must demonstrate in drills its ability to sustain 72-hour operations without IT systems. Hospitals use paper records, utilities switch to manual controls, and banks activate backup settlement channels.

Offline Backup Isolation: Maintain offline copies of control system configurations and verify the ability to rebuild IT systems from bare-metal states to ensure rapid restoration.

Resilient Emergency Preparedness: Pre-position regulatory waivers and authorization documents to rapidly deploy critical resources during major operational disruptions.

Scale Defense via AI (Operated Through ISACs)

Vulnerability Discovery & Remediation: Conduct AI-driven vulnerability mining on open-source code, vendor code, and firmware prior to release. Findings are relayed to manufacturers via the National Vulnerability Exchange Hub. Vendors must provide patches or alternative mitigations within strict deadlines, reserving dedicated rollout windows for critical infrastructure to deploy defenses before wide-spread vulnerability disclosure.

Risk Mitigation via Alternatives: For unpatchable systems, AI converts each vulnerability into specific compensatory controls (e.g., firewall rules, monitoring alerts, or configuration changes), which are rigorously tested in digital twin environments before deployment.

Intercept Malicious Traffic: CDNs, firewall providers, telecom operators, and network service providers can redirect and intercept malicious traffic before it reaches operational systems. Scaling these measures may require government authorization and support.

Provide Monitoring for Underserved Operators: Offer subsidized AI security operations services for small water systems, rural hospitals, and power cooperatives (nonprofit funding is currently available to support this initiative).

Conduct Realistic Attack Exercises: Run AI-driven red team exercises against sector digital twins, enabling defenders to practice countering attacker-speed tooling.

Safety Guardrails: AI defense tools default to read-only access for control networks; all changes require manual approval.

Sector-Specific Playbooks

Each sector has a primary risk, a priority plan, an executing ISAC, and a curated shortlist of CEOs selected based on company scale and industry influence.

Policy, Funding, and Metrics

Cross-Sector Members

All sectors operate on top of mainstream cloud providers (Microsoft, AWS, Google) and increasingly rely on frontier AI labs. These organizations will join the SIF CEO Committee as members, tasked with supplying AI defense tools and mining vulnerabilities in vendor products.

Policy Levers, Funding Support, and Proposed Timeline

The White House Super Intelligence Force must address policy, liability, and regulatory issues to incentivize necessary participation. The following key points outline priority institutional reforms requiring immediate advancement.

Measure Outcomes, Not Paperwork: Regulators track zero exposure for control systems, time-to-contain critical vulnerabilities, and validated manual operation durations, rather than auditing checklist compliance.

Performance-Based Regulatory Relief: Companies achieving outcome-oriented goals and securing board certification will receive safe harbor legal protections.

Enforce Vendor Accountability: Incorporate "secure by default" requirements into federal procurement; vendors must disclose product support lifespans and deliver patches or alternative mitigations for AI-discovered vulnerabilities within strict deadlines.

Information Sharing Guarantees: Drive long-term reauthorization of the Cybersecurity Information Sharing Act of 2015 (CISA 2015), enabling operators to rapidly share AI-generated threat intelligence under legal protection.

Defense Compute Assurance: The federal government coordinates access and compute support for frontier models for participating companies and the National Vulnerability Information Exchange Hub.

Emergency Response Talent Reserve: Establish a vetted emergency response reserve drawn from major operators and vendors, deployable to the most critical critical infrastructure sites during security incidents.

Timeline and Metrics

Core targets must complete risk containment within 6 months. While total replacement of unsupported legacy equipment was once proposed, its high cost and potential 2-to-5-year timeline necessitate prioritizing containment measures.

Addressing nationwide technical debt will likely entail massive costs, which must be strategically prioritized toward operators and shared suppliers whose failures would trigger nationwide cascading outages and lack unilateral capacity to upgrade system resources.