VC & Startups/Nov 15 edition/Reporting & analysis
Venture fundsEcosystems

Venture fundsFund closes, LPs & managers

a16z's Anne Neuberger proposes CEO pledges and 72-hour manual-operation drills for US critical infrastructure under the Super Intelligence Force

Anne Neuberger, an a16z general partner and former White House cyber official, proposes that the new Super Intelligence Force coordinate critical-infrastructure defense through CEO commitments: no internet-exposed control systems, proven 72-hour manual operation, and fixed deadlines for containing critical flaws.

From the a16z article of November 12, 2026 and cited sources: The plan covers five sectors. [1]

SOURCES[1] Super Intelligence-Era Critical Infrastructure Cyber… a16z.news

AI-assisted, not human-edited. Methodology Corrections Who owns this site

KEY FACTS
What happened
The plan covers five sectors. [1]
Original source
a16z article, independent perspective
Brief by
AT Pulse research desk AI-assisted; publisher and contact on the desk page
When
November 12, 2026 article published
Evidence
8 sources across 8 sites
61/100StrongHow it’s measured

Substack · 150 likes · 1 comment

Reach blends public engagement (views, likes, comments, shares, stars) from the platforms where the story or its original source lives: YouTube, Hacker News, GitHub, Substack and similar. A video and the article written from its transcript share one score, taken from the video. Platforms with no data are left out rather than counted as zero, and the age of a story never lowers its score.

Illustration from a16z: a16z's Anne Neuberger proposes CEO pledges and 72-hour manual-operation drills for US critical infrastructure under the Super Intelligence Force
Image: a16z — Original article ↗
THE CORE IDEAS4 TAKEAWAYS
01

The plan covers five sectors. It aims to contain the top few hundred operators and shared vendors within six months, compared with two to five years to replace legacy equipment. Both timeframes are the plan's own estimates, with no costing behind them. Operators would be ranked by the harm their failure would cause and by how many others depend on them, not by size. Shared vendors are treated as a key concentration of risk. [1]

02

On the technical side, AI would turn each flaw into a compensating control, such as a firewall rule, an alert or a configuration change, tested first on a digital twin. AI tools would be read-only on control networks by default, and a human would approve every change. Findings would flow through the Gold Eagle clearinghouse, widened to cover industrial equipment. Neither the plan nor Gold Eagle describes evaluation methods, error rates or OT-safe patching criteria. [1] [2] [3]

03

The legal and institutional footing is uncertain. The plan relies on liability protections in the 2015 information-sharing law, which now runs only to December 11, 2026, under a short-term extension. The Super Intelligence Force is a 120-day task force with a broad AI mandate, and its leaders have not been given a cyber-defense coordination role. [4] [5] [6]

04

Outside data support the urgency but shift the emphasis. Mandiant measured the handoff from initial access to the next attack stage falling from over eight hours to 22 seconds. Median patch time rose from 32 to 43 days. That suggests fixing flaws, more than finding them, is the main bottleneck. [2]

WHY IT MATTERS

attacker handoffs are speeding up while patching slows. Implication: operators and ICS vendors could face outcome-based scorecards and deadlines to fix flaws.

Read the full assessment

That depends on government adoption, funding, enforcement and lasting liability protection, and none of these is yet in place.

Sources are gathered; writing the dossier now.

Executive brief

A venture firm has published a plan for the White House's new "Super Intelligence Force" (SIF) to coordinate cyber defense of US critical infrastructure. Anne Neuberger, an a16z general partner and former Deputy National Security Advisor for cyber, wrote it. The plan depends on legal protections that are still unsettled: the information-sharing law it relies on now expires December 11, 2026, under a short-term extension.

Read the full section

A venture firm has published a plan for the White House's new "Super Intelligence Force" (SIF) to coordinate cyber defense of US critical infrastructure. Anne Neuberger, an a16z general partner and former Deputy National Security Advisor for cyber, wrote it. It names five sectors and would have CEOs sign commitments to three targets: no internet-exposed control systems, drills proving 72 hours of manual operation, and set deadlines for containing critical flaws. The goal is to contain the top few hundred operators and shared vendors within six months. The plan depends on legal protections that are still unsettled: the information-sharing law it relies on now expires December 11, 2026, under a short-term extension. It includes no budget and no independent analysis.

What changed and event timeline

  1. Executive order creates Gold Eagle

    EO 14409 set up a federal clearinghouse for AI-found vulnerabilities. The a16z plan wants this channel to carry industrial-equipment findings too ().

  2. Gold Eagle launches

    It runs on Carnegie Mellon SEI's VINCE platform and focuses on open-source software. Anthropic confirmed it is taking part; other participants were not named ().

  3. CISA 2015 gets another short extension

    The law's liability protections now run to December 11, 2026, not a long-term renewal ().

  4. Trump creates the Super Intelligence Force

    The task force is led by DNI Jay Clayton, named AI czar, and has a 120-day mandate to report on AI risks and benefits (;).

  5. Neuberger launches SAIF

    SAIF is a 501(c)(3) nonprofit that brings AI and cyber tools to under-resourced public institutions, such as the 1,800 rural hospitals. It starts with two cities and one sector ().

  6. The plan is published

    It came out on a16z's newsletter alongside a Wall Street Journal op-ed ().

Capabilities and access

This is a policy proposal, not a product. It names no model or version. The plan: All of this is in the plan.

Read the full section

This is a policy proposal, not a product. It names no model or version. The plan:

  • Asks the federal government to secure frontier-model access and compute for participating companies and the clearinghouse.
  • Makes Microsoft, AWS, Google and frontier AI labs members of the CEO council, responsible for AI defense tools.
  • Calls for a subsidized AI security-operations service for small water systems, rural hospitals and electric co-ops, and notes nonprofit funding is already available.

All of this is in the plan.

Technical analysis for researchers and developers

The plan's technical steps (a16z News): The plan does not describe evaluation methods, error rates or how digital twins would be checked for accuracy. Gold Eagle has published no criteria for deploying patches safely on OT systems as opposed to ordinary IT (CSA).

Read the full section

The plan's technical steps (a16z News):

  • Ranking: prioritize operators by how much harm their failure causes and how many others depend on them, not by size. Shared vendors are treated as often riskier than any single operator.
  • Containment: find exposed systems with AI-assisted scanning, use one-way data diodes, separate IT from control networks, keep offline configuration backups and prove systems can be rebuilt from bare metal.
  • AI workarounds: turn each flaw into a compensating control (a firewall rule, a monitoring alert or a configuration change), tested on a digital twin first.
  • Guardrails: AI tools are read-only on control networks by default, and a human approves every change.

The plan does not describe evaluation methods, error rates or how digital twins would be checked for accuracy. Gold Eagle has published no criteria for deploying patches safely on OT systems as opposed to ordinary IT (CSA).

Claims and evidence

  • AI is shrinking the time from flaw discovery to exploitation. This is the author's claim.
  • Remediation, not discovery, is the bottleneck.
  • Six-month containment, versus 2–5 years to replace old equipment. These are the plan's own estimates and have no costing behind them (a16z News).
Read the full section
  • AI is shrinking the time from flaw discovery to exploitation. This is the author's claim. Other data points the same way: Mandiant measured the handoff from initial access to the next attack stage falling from over 8 hours (2022) to 22 seconds (2025). Verizon's 2026 report has vulnerability exploitation causing 31% of breaches (CSA summary).
  • Remediation, not discovery, is the bottleneck. This cuts against the plan's emphasis on finding flaws: median patch time rose from 32 to 43 days (CSA).
  • Six-month containment, versus 2–5 years to replace old equipment. These are the plan's own estimates and have no costing behind them (a16z News).
  • Only the power and finance ISACs have CEO-level involvement. This comes from the author and was not independently confirmed in the reviewed sources.

Context and prior work

  • The plan builds on sector coordinating councils and ISACs (two-decade-old industry information-sharing bodies), the grid's GridEx attack exercise, the electricity sector's Cyber Mutual Assistance program and banking's Sheltered Harbor data-restoration standard (a16z News).
  • Gold Eagle overlaps with industry efforts such as the Linux Foundation's Akrites and Chainguard's Athena (Cybersecurity Dive).
  • The SIF is a temporary task force with a broad AI mandate. Its leadership was not named as having a cyber-defense coordination role (NPR).
Read the full section
  • The plan builds on sector coordinating councils and ISACs (two-decade-old industry information-sharing bodies), the grid's GridEx attack exercise, the electricity sector's Cyber Mutual Assistance program and banking's Sheltered Harbor data-restoration standard (a16z News).
  • Gold Eagle overlaps with industry efforts such as the Linux Foundation's Akrites and Chainguard's Athena (Cybersecurity Dive).
  • The SIF is a temporary task force with a broad AI mandate. Its leadership was not named as having a cyber-defense coordination role (NPR).

Limitations, safety and contested findings

  • Missing basics: the plan has no budget, no implementation schedule beyond its targets and no named operators.
  • Conflict of interest: a16z invests in AI and security firms that could take part, and the plan describes no safeguards (1023 Jack).
  • Target-list risk: a central store of unpatched-system status could become a ready-made target list for attackers if breached (CSA).
Read the full section
  • Missing basics: the plan has no budget, no implementation schedule beyond its targets and no named operators. It does not explain how the SIF would enforce anything (1023 Jack analysis).
  • Conflict of interest: a16z invests in AI and security firms that could take part, and the plan describes no safeguards (1023 Jack).
  • Target-list risk: a central store of unpatched-system status could become a ready-made target list for attackers if breached (CSA).
  • Weak legal footing: liability cover rests on CISA 2015, which is on short-term extension (CRS).

Business and practitioner implications

  • OT operators: the proposed scorecard measures exposure, time to contain a flaw and hours of manual operation, not compliance checklists.
  • ICS vendors: they could face deadlines to fix flaws as a condition of selling to critical operators, along with requirements to disclose how long they will support products.
  • Cloud, CDN and telecom providers: they may be asked to block malicious traffic at national scale, which could require new legal authorities.
Read the full section
  • OT operators: the proposed scorecard measures exposure, time to contain a flaw and hours of manual operation, not compliance checklists. For public companies, board attestation would replace duplicate audits (a16z News).
  • ICS vendors: they could face deadlines to fix flaws as a condition of selling to critical operators, along with requirements to disclose how long they will support products.
  • Cloud, CDN and telecom providers: they may be asked to block malicious traffic at national scale, which could require new legal authorities.
  • Overall: this is commentary from an investor-affiliated author. No government adoption has been reported.
FOLLOW THE EVIDENCE

The source trail.

Sources (8)
A LITTLE LESS NOISE. A LOT MORE CONTEXT.

Stay curious.
Follow the evidence.

Independent perspectives, the original sources, and room for the questions that don't have easy answers.

How we build the brief
Connect with us

Find us where you already read.