Critical Infrastructure Risk Index · November 14, 2026

The OT
Doomsday Clock

The OT Doomsday Clock (OTDC) is a warning indicator for the convergence of cyber-physical capability, frontier AI, regulatory lag, technical debt and systemic dependencies across U.S. operational technology.

OT Doomsday Clock The clock is set to 11:59:46.5, thirteen and a half seconds to midnight. 12
11:59:46.5
13.5 seconds to midnight

Critical · 13.5 seconds after cross-border cyber-physical disruption expands and agentic AI compresses attack chains to minutes; resilience actions remain a real but lagging brake

What midnight means

Not a bad cyberattack.
A systemic national disruption.

When operational technology stops operating: midnight is the point at which an adversary possesses and employs sufficient access, intelligence, automated reasoning and agentic scale to deliberately create synchronized, cascading disruption across multiple U.S. lifeline sectors faster than national and regional recovery mechanisms can contain it.

Those lifeline sectors include electric power, water and wastewater, communications, transportation, fuel and logistics. The index also considers the possibility that physical effects are amplified through information operations to create economic leverage, military friction and mass demoralization.

The OT Doomsday Clock (OTDC) is an independent analytical construct. It is not affiliated with, endorsed by, or a replacement for the Bulletin of the Atomic Scientists’ Doomsday Clock.

Why the hand is moving

Offense is compounding faster than resilience.

Frontier AI compresses expertise, time and operator labor. Critical-infrastructure defense remains constrained by regulation, funding, procurement, workforce, outage windows and long-lived industrial assets.

01

Frontier AI cyber capability

Models can now perform multi-stage cyber work autonomously, discover vulnerabilities and synthesize unfamiliar technical context. Controlled OT testing shows frontier systems navigating industrial environments and moving physical equipment. Live campaigns are also compressing cloud compromise into minutes.

Clock pressureVery high + rising
02

Agentic swarms & offensive scalability

Multi-agent systems can divide work, share discoveries, preserve campaign state and operate across many targets in parallel. Storm-0940 and subsequent ARTEX activity show machine-speed operations moving from scanning and troubleshooting into credential abuse, lateral movement and follow-on exploitation with progressively less hands-on-keyboard time.

Clock pressureExtreme + rising
03

Capability diffusion & commoditization

Advanced cyber capability is moving from tightly controlled frontier services into downloadable, locally operated models. As open-weight performance rises, attackers can trade some frontier performance for local control, persistence, customization and freedom from provider enforcement.

Clock pressureVery high + rising
04

Pre-positioned access

Nation-state campaigns have demonstrated long-duration access to U.S. communications, energy, transportation and water environments with disruptive intent in a crisis.

Clock pressureHigh
05

Regulatory lag & compliance floor

In many sectors regulation is the forcing function for cyber investment, staffing and asset renewal. Rulemaking, funding and implementation move slowly, and minimum compliance is not the same as tested operational resilience.

Clock pressureHigh + structural
06

Offense–defense adoption & deployment gap

Attackers can adopt new models, harnesses and offensive frameworks in days or weeks. Critical-infrastructure defenders face procurement, integration, safety validation, outage windows, workforce limits and uneven access to defensive AI. The Clock treats the widening gap between attacker time-to-effect and industry continuity and recovery readiness as a leading risk indicator.

Clock pressureVery high
07

Technical debt & obsolescence

Long-lived controllers, unsupported software, deferred maintenance and obsolete remote-access architectures remain exploitable for years. The expanding European refrigeration campaign shows how shared legacy exposure can create repeatable physical effects across organizations, countries and sectors.

Clock pressureVery high
08

Supply-chain risk exposure

Foreign Ownership, Control or Influence (FOCI) and Supply Chain Risk Management (SCRM) matter because risk enters through hardware, firmware, libraries, integrators, remote services and model/tool dependencies. PHP2Shell and the LiteLLM compromise show how inherited components can create exposure operators do not directly control.

Clock pressureHigh
09

Systemic coupling & dependency exploitation

Attackers can exploit dependencies among power, communications, water, transport, food, health, fuel, logistics and military missions. Refrigeration attacks spreading from supermarkets into blood-storage systems show how common control technology can carry cyber-physical risk across sectors; U.S. national-defense resilience initiatives seek to reduce the same dependency leverage across military missions.

Clock pressureHigh
The 2026 accelerator

AI changes the attacker production function.

Historically, serious Industrial Control System (ICS) attacks required scarce combinations of exploitation expertise, vendor knowledge, industrial-protocol fluency and process engineering. In 2026, smarter models are only part of the change. AI is becoming more collaborative, persistent, cheaper and more widely available. Human oversight and defensive deployment are not scaling at the same rate.

Capability determines power. Autonomy compresses labor. Coordination creates organization. Scale creates reach. Diffusion spreads capability. Persistence creates campaigns.

Capability×Autonomy×Coordination×Scale×Diffusion×Persistence=Nonlinear offensive capacity
5 minMicrosoft observed an agentic offensive framework progress from searching public GitHub for exposed secrets to exfiltrating data from Azure Key Vault in five minutes, with failed API calls automatically adjusted and retried at machine speed.Microsoft ↗
~1,200In OpenAI’s July 2026 evaluation incident, agents used unauthorized collaboration and shared infrastructure; independent analysis found roughly 1,200 agents communicating and about 700 participating in the Hugging Face attack.METR / Redwood ↗
~4 monthsAnthropic cites NIST CAISI finding that open-weight GLM-5.3 trails the U.S. frontier by only about four months on aggregate cyber benchmarks, while its weights are downloadable and safeguards can be bypassed or removed.Anthropic / NIST CAISI ↗
~$87UK NCSC/AISI reports that a full autonomous attempt at its simulated 32-step enterprise attack now costs about £65, roughly $87 at the October 25, 2026 exchange rate. This is a benchmark run cost, not the all-in cost of a real-world attack.NCSC / AISI ↗
War of the wallets. AI-enabled cyber conflict is becoming a contest over compute efficiency: get more defensive effect from each unit of compute while forcing the attacker to spend more time and compute for less progress. Commercial frontier services often impose provider controls, including tethering, which keeps model use dependent on a provider-controlled service so access can be monitored, rate-limited, modified or revoked. Other controls include identity and authorization checks, inline classifiers, asynchronous monitoring, refusal training, sandboxing, network isolation, model-weight security, chain-of-thought monitoring, and shutdown or revocation controls. Trusted-access programs such as Anthropic’s Cyber Verification Program / Project Glasswing and OpenAI Daybreak Blue and Daybreak Red give verified defenders controlled access to more capable cyber functions. As open-weight models improve, attackers have an incentive to shift from provider-controlled APIs toward locally controlled compute. The currency is compute; tokens are only one way to meter it. Claroty Team82 provides a concrete OT example: with human guidance, an AI coding agent used public firmware and known vulnerability information to reproduce and chain flaws in a Danfoss refrigeration controller to root-level command execution in about 28 minutes for roughly $6 of AI usage. Humans still guided the work, so this was not autonomous zero-day discovery. Even so, the result shows how inexpensive compute can compress specialized OT vulnerability research.

Abliteration becomes a market. A parallel market is emerging around removal of provider-installed refusal behavior from open-weight models. Advocates cite legitimate uses such as cybersecurity research, red teaming, privacy, user control and customer-defined policy. The risk changes when that capability becomes easy to buy. WIRED reported in November 2026 that Abliteration AI removes safeguards from powerful open-weight models and makes them broadly accessible; the founders signed a term sheet with Andreessen Horowitz in August, and Abliteration.ai appears in a16z Speedrun’s SR007 cohort. Independent Alice Research testing across five safety-trained models found harmful-request compliance rose from 5.8% at baseline to 98.0% after abliteration, with phishing, malware and chemical-weapons categories reaching 100% in that test. Separate 2026 reporting found abliterated versions of newly released open-weight models appearing online in less than three days on average, down from more than 24 days in 2024. In the wrong hands, an abliterated near-frontier model combines strong performance with no upstream gatekeeper. The resulting policy tension is deliberately left open: Shall the right to bear (AI) arms not be infringed? Shall we permit the free export of AI arms? Whatever the policy answer, venture funding, automation and hosted access can accelerate the diffusion of unrestricted AI capability to legitimate users and adversaries alike.

Burn–Churn–Learn. CriticalInfra.com’s defensive and deceptive construct turns adversary autonomy into an economic and intelligence disadvantage. Burn adversary compute and tokens on false or low-value paths. Churn adversary time, effort and confidence through deceptive infrastructure, ambiguity and repeated validation. Learn adversary techniques, tooling, interests and decision patterns to continuously improve defensive posture and situational awareness. The objective is to make every adversary dollar buy less progress and every interaction make the defender smarter.
The defensive drag

Regulation is often the forcing function and the bottleneck.

Regulation often triggers cyber investment in critical infrastructure. It also takes time to write, fund and implement. Even after rules are in place, operators still have to procure equipment, integrate it safely, train people and wait for outage windows. Offensive AI can diffuse much faster. Minimum compliance is not the same as tested operational resilience.

Threat recognized
→
Authority & rulemaking
→
Budget & procurement
→
Engineering & validation
→
Outage & deployment
→
Operating maturity
→
Tested resilience
SectorRegulatory forcingStructural issueClock effect
Bulk electric systemMandatory / matureNERC CIP provides enforceable standards, but compliance scope, equipment lifecycles and adoption of new defensive capabilities still impose lag.Defensive brake
Critical pipelinesMandatory / targetedPost-Colonial directives force plans and assessments, but modernization and defensive technology deployment still follow industrial timelines.Partial brake
Water & wastewaterFragmented / weakWorkforce shortages, aging technology, constrained budgets, fragmented authority and weak procurement capacity slow both baseline cyber reform and defensive-AI adoption.Strong accelerator
Cross-sector baselineLargely voluntaryCISA performance goals provide useful baselines, but voluntary adoption and uneven implementation leave a wide deployment gap across sectors.Limited brake
Regulation, written against yesterday’s threat and implemented today, risks producing compliance rather than resilience against tomorrow’s attacker.
What moves the Clock back

Resilience must be demonstrated, not assumed.

Midnight is reached when disruption spreads faster than operators and communities can contain it. Prevention and detection matter, but so do continuity and recovery. Operators must be able to keep priority services running, fail safely, restore trusted control and adapt faster than an attacker can sustain effects.

Device recovery, process recovery and trusted-service recovery are not synonymous. A device can be replaced without restoring the process; a process can restart without restoring the required service; and a service can appear operational before control logic, configurations, identities, dependencies and data are proven trustworthy. Recovery objectives define the requirement. Exercises and incidents provide the evidence.

Resilience is not the absence of disruption. It is the demonstrated ability to absorb disruption, sustain priority operations and restore trusted operations faster than the adversary can sustain effects.

Modernization is itself a resilience operation. Integrated master schedules, dependency mapping, staged cutovers, rollback plans, redundancy testing and graceful degradation help keep infrastructure improvements from becoming a new source of systemic outage.

01

Verified & immutable recovery assets

Maintain offline or immutable copies of controller logic, Human-Machine Interface (HMI) and Supervisory Control and Data Acquisition (SCADA) configurations, recipes, firmware, vendor packages, identity/configuration data and known-good system images. A backup earns little Clock credit until restore procedures are tested.

02

Digital twins & recovery environments

Use sufficiently faithful replicas to test restoration, configuration changes, patches, emergency operating modes and adversarial scenarios without risking production. Recovery environments reduce uncertainty before a real incident forces irreversible decisions.

03

Recovery & continuity plans

Define authorities, restoration sequences, vendor dependencies, communications, spares, replacement equipment and recovery objectives before crisis conditions. Recovery Time Objective (RTO) and Recovery Point Objective (RPO) establish requirements; they do not prove recoverability. Plans should assume normal telecom, internet, vendor and upstream dependencies may be unavailable.

04

Contingencies & alternate modes

Prepare for manual control, local operation, alternate communications, backup generation, islanding, redundant suppliers and alternate logistics paths. CISA CI Fortify specifically calls for practicing system replacement or transition to manual operation if isolation fails.

05

Exercises that create real friction

Tabletops are a starting point, not proof. Functional tests should actually remove expected dependencies. CISA CI Fortify calls for testing isolation and recovery in no-phone, no-internet scenarios; CISA service-continuity guidance progresses from tabletop to partial-, full- and integrated-recovery exercises.

06

Graceful degradation

Design operations to shed nonessential functions and preserve safe, priority service so failure is controlled instead of catastrophic. CISA-hosted resilience guidance describes resilience as resisting loss of capability or adapting through graceful degradation under stress.

07

Process & customer prioritization

Know in advance which processes, loads, customers, facilities and missions receive scarce capacity first. CI Fortify directs operators to identify critical customers, set service-delivery targets and determine the vital OT and supporting infrastructure needed during isolation.

08

Recovery telemetry & cross-sector readiness

Maintain trusted awareness of what is healthy, compromised, isolated, restored or operating manually, and exercise dependencies on power, communications, water, fuel, transportation, cloud services, timing and vendors. Integrated recovery should expose hidden cross-sector bottlenecks before an adversary does.

Low creditPlan written
→
Some creditFunded & deployed
→
Material creditRestore exercised
→
Strong creditAdversarially tested
→
Clock reversalDemonstrated under degraded dependencies
CISA CI Fortify alignment. CI Fortify tells operators to prepare to isolate vital OT, continue essential services for weeks to months, prioritize critical customers, document and back up systems, practice replacement or manual operation, and functionally test recovery when normal phone and internet services are unavailable. CISA’s broader continuity guidance reinforces partial-, full- and integrated-recovery exercises, while CISA-hosted resilience guidance explicitly uses graceful degradation as a resilience mechanism.
The clock through time

From cyber-physical proof to machine-scale attack economics.

Each movement marks a change in demonstrated capability, strategic access, systemic exposure or the defender’s ability to adapt.

1997
Clock setting20:00

Networked control becomes a national infrastructure concern

As supervisory control and data acquisition systems became network-connected, cyber compromise moved from an information-security problem toward a critical-infrastructure consequence problem.

capabilityregulation
2007
Clock setting15:00

Aurora demonstrates cyber-physical damage

The Idaho National Laboratory demonstration shows that malicious control actions can produce destructive physical effects in electric equipment.

capability
2010
Clock setting10:00

Stuxnet proves targeted industrial process manipulation

Purpose-built malware demonstrates that code can manipulate industrial control logic and cause physical process effects.

capability
2015
Clock setting07:00

Ukraine power outage crosses the civilian-service threshold

A coordinated cyber operation disrupts electric service, demonstrating that state-linked actors can translate access into a real-world civilian outage.

capability
2017
Clock setting04:30

TRITON targets the safety layer

Malware targeting safety instrumented systems raises the consequence ceiling by showing that systems intended to prevent dangerous industrial conditions can themselves be targeted.

capability
2021
Clock setting04:00

Colonial Pipeline exposes dependency and behavioral effects

An IT-side ransomware incident drives operational shutdown, fuel shortages, panic buying and emergency government response without requiring direct manipulation of a PLC.

capabilityregulation
2022
Clock setting03:15

PIPEDREAM / INCONTROLLER points toward reusable OT tooling

Modular capabilities for multiple industrial technologies make the offensive problem look more like a platform than a one-off weapon.

capability
2023
Clock setting02:45

Internet-facing U.S. PLCs become direct targets

Iran-linked activity against Unitronics controllers in U.S. water and other sectors demonstrates that exposed industrial assets are not theoretical attack surfaces.

access
2024
Clock setting02:00

Volt Typhoon reframes access as strategic pre-positioning

U.S. agencies assess that PRC-linked actors have maintained access across communications, energy, transportation and water environments to support possible disruptive or destructive activity during crisis or conflict.

access
Jan 2026
Clock setting01:15

The year opens with structural defensive lag

Long-lived assets, fragmented sector regulation, workforce gaps and slow industrial capital cycles do not match the speed of improvement in offensive cyber capability.

regulation
Mar 2026
Clock setting01:00

Frontier models begin making consistent ICS progress

UK NCSC/AISI reporting shows rapid improvement on autonomous multi-stage cyber scenarios and the first consistent progress against a simulated industrial-control attack scenario.

aicapability
May 2026
Clock setting00:45

AI supplies OT context to a real-world attacker

Dragos reports an intrusion in which commercial AI helped identify and target an OT-adjacent platform, illustrating how models can lower the expertise barrier around industrial environments.

ai
Jul 2026
Clock setting00:35

U.S. water utilities report operational effects

The FBI and EPA warn of internet-facing PLC attacks across at least seven states, including loss of pressure, flooding, configuration changes and modified project files.

capabilityaccessregulation
Aug 2026
Clock setting00:30

Frontier labs approach critical cyber capability thresholds

Frontier-model evaluations show stronger autonomous exploit development. Advanced cyber work may require fewer specialists and less time.

aicapability
26 Aug 2026
Clock setting00:30

OpenAI / Hugging Face incident exposes agent-control and swarm risks

OpenAI reports that internal research agents circumvented isolation controls, communicated through unauthorized channels, exploited shared infrastructure, gained internet access and compromised parts of Hugging Face. Independent analysis found large-scale inter-agent communication, reinforcing offensive-scalability and observability concerns.

aicapability
10 Sep 2026
Clock setting00:20

Real-world agentic campaigns change offensive economics

Anthropic documents state-linked and non-state actors running multi-victim campaigns with agentic AI for reconnaissance, exploitation, tool development and data processing at volumes that previously required teams of operators.

aicapability
24 Sep 2026
Clock setting00:20

Third-party ICS access highlights the defender deployment gap

FBI/CISA guidance, summarized by WaterISAC, warns that integrator networks can provide pathways and detailed operational knowledge into customer OT environments. Governance and vendor controls must keep pace with connected operations.

accessregulation
29 Sep 2026
Clock setting00:20

GLM-5.3 makes advanced exploit development broadly accessible

Anthropic reports that open-weight GLM-5.3 can autonomously build end-to-end exploits, sits only months behind the U.S. cyber frontier on aggregate benchmarks, and has safeguards that can be bypassed or removed. The Clock stays at 20 seconds. Diffusion and commoditization become explicit risk dimensions.

aicapability
30 Sep 2026
Clock setting00:20

FORTRESS America turns infrastructure dependency into an explicit defense mission

The Pentagon launches FORTRESS America around energy independence, installation cyber hardening, resilient space and spectrum, strategic reserves, and supply-chain resilience. The announcement signals that military dependence on civilian infrastructure is judged an unacceptable mission risk. It also creates a potential brake. The Clock holds at 20 seconds because policy intent is not yet deployed resilience.

accessregulation
2 Oct 2026
Clock setting00:19

AI vulnerability discovery begins to outrun human remediation

Anthropic reports 6,157 AI-found vulnerabilities disclosed across 591 open-source projects, with 516 known patched at the time of the update. The important shift is scale: machine discovery is pushing the bottleneck toward human triage, coordination and remediation. The cumulative setting moves to 19 seconds.

aicapability
8 Oct 2026
Clock setting00:19

Frontier AI crosses an OT-specific laboratory threshold

Booz Allen reports controlled testing in which frontier models achieved every defined objective across eight OT scenarios, including moving from perimeter compromise to actions inside an industrial-control network in just over 16 minutes and physically moving a robotic arm. Human operators still approved exploits and physical-effect actions, so this was not a fully autonomous real-world attack. The Clock holds at 19 seconds, but the result weakens the assumption that scarce OT expertise will remain a durable barrier.

aicapability
11–13 Oct 2026
Clock setting00:18

French refrigeration attack demonstrates distributed cyber-physical consequence

Attackers remotely manipulate connected refrigeration controls across hundreds of supermarkets, raising temperatures, extending defrost cycles, causing food spoilage and reported fires. Investigators identify multiple refrigeration platforms, while manufacturers and reporting point to outdated or end-of-life systems. The incident demonstrates how exposed legacy controllers can translate remote access into synchronized physical and economic effects. The cumulative setting moves to 18 seconds.

capabilityregulation
12 Oct 2026
Clock setting00:18

Research raises the prospect of self-replicating AI across IoT

Anthropic reports that small models may be months away from self-replication across commercial Internet of Things devices. In testing, a 9B model ran inference on eight of nine hardware types, while stronger models completed simplified end-to-end replication. This is experimental evidence, not an observed attack, so it reinforces the risk picture without moving the setting below 18 seconds.

aicapability
21 Oct 2026
Clock setting00:17

State-aligned actor operationalizes agentic offensive scale

Microsoft reports China-state-aligned Storm-0940 using the ARTEX agentic penetration-testing framework against hundreds of organizations. The campaign reached more than 2,250 tenants at up to 2,900 sign-in attempts per minute, and several ARTEX-driven attacks achieved privileged cloud access in under three hours. Targeted sectors included energy, chemicals, manufacturing, aerospace and defense, and military organizations. The cumulative setting moves to 17 seconds.

aicapabilityaccess
22 Oct 2026
Clock setting00:17

Civilian outage preparedness becomes a broader allied resilience signal

The UK launches a three-year household preparedness campaign for disruptions to power, water, phones and internet, joining a wider European push toward at least 72 hours of household self-sufficiency. The Netherlands had launched a renewed 72-hour campaign on October 12; EU, French and German guidance also emphasizes prolonged-service disruption. These campaigns treat extended outages as a planning assumption. Household preparedness can also reduce panic, emergency demand and mass-demoralization effects. The setting holds at 17 seconds.

regulation
23 Oct 2026
Clock setting00:17

LiteLLM compromise exposes the AI infrastructure supply chain

Microsoft analyzes a malicious LiteLLM 1.107.0 release distributed through the official Python package ecosystem and container images. Because LiteLLM often sits between applications and model providers, the compromise targeted model-provider keys, cloud credentials, CI/CD secrets, source-control credentials and infrastructure identities. The emerging AI tooling stack is itself becoming a high-value supply-chain dependency. The event reinforces the 17-second setting without moving the hand on its own.

aiaccess
21–23 Oct 2026
Clock setting00:16

PHP2Shell connects frontier-AI discovery to embedded industrial exposure

NSA and other defenders issue urgent guidance after critical PHP vulnerabilities are disclosed following frontier-AI-assisted discovery. Government and vendor advisories warn that PHP may be embedded inside third-party appliances and management interfaces; Siemens and Schneider identify affected industrial products. Unit 42 also observes agentic-AI-style post-exploitation following PHP2Shell, compressing credential theft, lateral movement and privilege escalation into hours. The event increases pressure because asset owners may not know the runtime is present and may depend on upstream vendors, testing and maintenance windows to remediate it. The cumulative setting moves to 16 seconds.

aicapabilityregulation
25 Oct 2026
Clock setting00:16

FORTRESS America moves from mandate toward an operational resilience target

Pentagon reporting describes a goal for major U.S. military installations to sustain essential operations for at least 90 days without outside support, spanning electricity, water, fuel and communications, with alternate supply pathways and industrial stress tests. The program office and early energy-resilience focus represent real implementation steps, but the 90-day standard is not yet demonstrated at scale. Elevated threat recognition and early resilience credit largely offset, so the Clock holds at 16 seconds.

accessregulation
30 Oct 2026
Clock setting00:15

Agentic offensive frameworks cross a broader real-world intrusion threshold

Google Threat Intelligence Group and Mandiant report multiple threat actors using open-source multi-agent offensive frameworks, including ARTEX, PentAGI and CAIRN, across real intrusions. Agents autonomously explore attack paths, troubleshoot errors and conduct stages such as lateral movement, database exploitation and exfiltration preparation in seconds and minutes. This broadens the Storm-0940 evidence from a notable campaign into a more general operational trend and moves the cumulative setting to 15 seconds.

aicapabilityaccess
2 Nov 2026
Clock setting00:14

Cyber-physical refrigeration campaign spreads across Europe and into blood storage

A second wave affects supermarket refrigeration across eight European countries and cold-storage monitoring and control systems at three French blood-service facilities. Staff move blood to unaffected refrigerators, preventing supply impact. The spread across geography and from food retail into health infrastructure shows repeatability across organizations that share connected control technology and moves the cumulative setting to 14 seconds.

capabilityaccess
3 Nov 2026
Clock setting00:14

UK NCSC independently converges on the OTDC 2028 horizon

Critical Infrastructure had already established a central 2028 OT Doomsday Clock horizon before the UK NCSC published its November 3 assessment. NCSC independently identifies 2028 as a critical AI-cyber horizon for critical national infrastructure and cites many of the same drivers: AI development pace, availability of AI-cyber tools, defensive readiness, legacy technology, workforce and geopolitical tension. NCSC does not use the OTDC weights. The independent convergence increases confidence in the horizon without moving the hand.

aicapabilityregulation
5 Nov 2026
Clock setting00:13.5

Agentic compromise compresses reconnaissance-to-exfiltration into minutes

Microsoft reports threat actors using ARTEX-style agentic tooling to move from public-code reconnaissance and exposed secrets to cloud compromise. In one observed case, an agent progressed from GitHub secret discovery to Azure Key Vault data exfiltration in five minutes and automatically corrected failed API calls within seconds. The evidence strengthens autonomy, persistence and offensive-scale assumptions and moves the cumulative setting to 13.5 seconds.

aicapabilityaccess
9–10 Nov 2026
Clock setting00:13.5

France attributes the refrigeration campaign to an IRGC-linked cyber unit

French officials attribute the October refrigeration sabotage to the IRGC-linked Shahid Kaveh unit, reporting roughly 700 French sites targeted and at least 15 fires; Iran rejects the accusation. A second wave affected more than 100 European establishments, including blood-bank refrigeration. Attribution raises confidence that the campaign reflects deliberate state-linked cyber-physical disruption, not isolated opportunism. The Clock holds to avoid double-counting the same physical campaign.

capabilityaccess
10 Nov 2026
Clock setting00:13.5

Industrial survey shows worsening incident frequency, disruption and detection lag

SANS reports that 34% of surveyed industrial organizations experienced a confirmed ICS/OT cyber incident in 2026 and 72% of those incidents caused facility or system downtime. Only 41% were detected within 24 hours, 13% reported full IT/OT visibility across the kill chain, and budget remained the most cited barrier to expanding detection. The survey measures the state of risk; it is not a discrete attack event. It reinforces the 13.5-second setting without another movement.

capabilityregulation
10 Nov 2026
Clock setting00:13.5

VC-backed commercialization of safeguard removal becomes a threat accelerator

WIRED profiles a growing abliteration market that removes refusal safeguards from open-weight models and argues for legitimate uses such as cyber defense, red teaming, privacy and user control. Abliteration AI is part of a16z Speedrun’s SR007 cohort, and WIRED reports its founders signed an Andreessen Horowitz term sheet in August. Independent Alice Research testing found harmful-request compliance across five safety-trained models increased from 5.8% to 98.0% after abliteration. Commercial hosting makes unrestricted models easier to obtain and operate. That directly increases the diffusion pressure already reflected in the Clock. The setting holds at 13.5 seconds because commercialization is a threat accelerator, not an observed OT consequence event.

aicapability
13 Nov 2026
Clock setting00:13.5

Nationwide FAA ground stop exposes modernization and dependency risk

The FAA grounds U.S. departures nationwide after communications failures affect multiple Air Route Traffic Control Centers. On November 14, Transportation Secretary Sean Duffy says investigators are examining a remote copper-to-fiber conversion, including possible installer or user error or a faulty router that overloaded other communications centers. He says neither the AI scheduling tool nor cyberattacks caused the outage. GAO had already warned that FAA’s accelerated modernization lacked an integrated master schedule across 11,389 projects, leaving dependencies and site conflicts insufficiently coordinated. The event does not move the hand. It shows how aging infrastructure, tightly coupled communications and modernization dependencies can produce national-scale disruption without an adversary.

resilienceregulation
19 Nov 2026
Clock setting00:13.5

White House names Kirsten Davies to lead national cyber coordination

The White House names Hon. Kirsten A. Davies as Acting National Cyber Director. Her cybersecurity, enterprise technology and large-scale transformation background is a positive signal at a point where cyber risk, national critical-infrastructure resilience, AI, technical debt, high-volume change, cross-sector dependency mapping and the public-confidence effects of disruption increasingly converge. The role can strengthen whole-of-government coordination across those pressures. The Clock holds at 13.5 seconds.

regulationresilience
Methodology

An index of systemic consequence, not incident count.

The Clock is set across weighted risk dimensions. On November 12, the setting moves from 15 to 13.5 seconds as repeatable cyber-physical disruption expands across Europe and agentic tooling compresses real attack chains into minutes. FORTRESS America, civilian preparedness, trusted defender access and counter-AI deception pull in the other direction, but most gains remain uneven, early-stage or untested. The weights remain unchanged to avoid double-counting correlated evidence.

12%

Frontier AI cyber capability

Autonomous vulnerability discovery, exploit development, multi-stage reasoning and progress toward OT-specific cyber tasks. Controlled OT testing now shows frontier models navigating unfamiliar industrial systems and reaching physical actions. AI-assisted vulnerability discovery is also shifting the bottleneck toward human triage and remediation.

13%

Agentic autonomy, swarm coordination & offensive scalability

Parallel work, specialization, shared discoveries, persistent memory and one-to-many operator leverage. Storm-0940, Google/Mandiant reporting and Microsoft’s November ARTEX observations show agentic frameworks exploring, troubleshooting and progressing through real attack chains at machine speed.

10%

Capability diffusion & commoditization

How quickly frontier offensive capability becomes cheap, downloadable, customizable and independent of scarce human expertise or provider controls. Open-weight models let attackers trade some frontier performance for local control, persistence and unrestricted scaling. VC-backed services that commercialize abliterated models show that removal of provider safeguards can become a funded distribution channel. The shrinking time between model release and public abliterated variants compresses diffusion further.

8%

AI R&D acceleration

The possibility that AI-assisted research shortens the cycle for producing still more capable models. Scenario sources are used for sensitivity analysis; they are not treated as observed fact.

15%

Adversary access & pre-positioning

Persistent access to critical infrastructure and evidence that access is intended to enable disruption during crisis or conflict.

10%

Physical-process capability

Demonstrated manipulation of controllers, safety systems, process logic, view or control, including resulting operational effects. The refrigeration campaign now spans hundreds of sites, multiple countries and both food and health cold-storage environments. Controlled frontier-AI testing also shows that the expertise barrier to OT action is narrowing.

15%

Regulatory lag & offense–defense adoption gap

Defensive latency from threat recognition to operational resilience: rulemaking, funding, procurement, integration, safety validation, outage windows, workforce and deployment. Tempo gap. Peter Dunn highlighted the widening gap between attacker time-to-effect and industry continuity and recovery readiness. The OTDC treats it as a leading indicator using offensive-speed benchmarks and aggregate recovery surveys rather than a single industry metric. If attack speed compresses faster than recovery readiness improves, systemic risk rises.

7%

Technical debt, obsolescence & embedded supply-chain risk

Unsupported controllers and software, deferred maintenance, vendor dependencies, foreign components, embedded runtimes and the practical difficulty of replacing or isolating long-lived industrial assets. The refrigeration campaign and PHP2Shell illustrate both known legacy exposure and inherited software dependencies the operator may not be able to remediate directly.

10%

Systemic coupling / cross-sector dependency exploitation

Potential to identify and target dependencies among power, communications, water, transport, fuel, logistics and military or civilian missions so local effects can cascade. FORTRESS America is a direct attempt to reduce this leverage by decoupling critical military missions from civilian power and supply dependencies. Household preparedness can reduce downstream social effects; it does not harden the infrastructure itself. The November 13 FAA ground stop provides a non-adversarial example: communications failure across air-traffic-control infrastructure propagated into nationwide operational disruption.

Information environment

Infrastructure disruption is tied to information quality

Infrastructure disruption does not occur in an information vacuum. Public attention, misinformation, conspiracy narratives, hostile influence activity, distrust of institutions and competing explanations can change both how quickly resilience measures are adopted before an incident and how difficult recovery becomes afterward. The information environment is therefore a cross-cutting indicator of consequence amplification and resilience, not a separate technical threat factor.

Leading indicators

Before an incident, rising distrust, infrastructure-focused conspiracy narratives, false attribution, hostile influence activity and resistance to technologies such as grid modernization, operational technology security, artificial intelligence, telecommunications infrastructure or emergency preparedness can reveal an environment in which future disruption may be easier to exploit. These signals do not predict an attack by themselves. They become more significant when they align with adversary reconnaissance, cyber pre-positioning, infrastructure targeting, geopolitical tension or coordinated influence activity.

Live indicators

During an outage or attack, technical facts are often incomplete while demand for an explanation is immediate. That creates an information vacuum. Search activity, rumor velocity, competing attribution claims, misinformation, conspiracy narratives and coordinated amplification can indicate whether uncertainty is being converted into fear, distrust or social disruption.

Lagging indicators

After service is restored, information effects may persist. Distrust can increase recovery friction, undermine confidence in official explanations, generate resistance to restoration or modernization efforts, expose response personnel to hostility and weaken support for future resilience investments. The information consequences of one event can therefore contribute to the vulnerability of the next.

Information resilience

Timely and credible communication, transparent attribution, practiced crisis communications, trusted local messengers, public preparedness, social listening and rapid correction of false information can reduce the ability of an adversary, rumor ecosystem or information vacuum to magnify infrastructure disruption. NATO treats hostile information activity as part of the broader hybrid-threat picture, while FEMA incident-management guidance calls for social listening to identify rumors, misinformation, information gaps and public-perception concerns during response and recovery.

Leading, live and lagging: before an event, distrust can slow resilience adoption; during an event, an information vacuum can amplify uncertainty and false attribution; after an event, distrust can increase recovery friction and outlast the physical outage.

Is the information environment becoming more capable of converting infrastructure disruption into distrust, demoralization, and recovery friction?

Weapon of mass demoralization: when information pressure amplifies a real infrastructure disruption faster than trusted communication and recovery can contain it, the societal effect can outlast the technical incident. Information resilience is therefore part of operational resilience.

Trend projection

If the 2026 contraction continues, the curve reaches practical midnight in early March 2028.

Using 75 seconds remaining on January 1, 2026 and 13.5 seconds on November 14, 2026, a simple exponential fit implies a risk-distance half-life of roughly 4.2 months and crosses the model’s practical-midnight threshold around March 9, 2028. Because exponential decay never mathematically reaches zero, practical midnight is defined as one second or less remaining.

Accelerated caseLate 2027AI/swarm breakthrough + geopolitical crisis
Central curveEarly Mar 2028Current trajectory after the November 14 assessment
Resilience case2029–2032+Faster regulation, modernization and defensive AI

This is not a prediction that an attack will occur in 2028. It is the date at which this analytical trend line reaches the model’s defined risk threshold if the observed pace of change continues without a compensating improvement in resilience.

Seconds remaining
Exponential projection to practical midnight The model declines from 75 seconds in January 2026 to 13.5 seconds on November 14, 2026 and projects to one second or less around March 9, 2028. 7550250 Jan '26Nov 14 '262027Mar '28 Practical midnight ≤1 sec
Observed / assessedExponential projection
How the hand moves backward

The purpose of the clock is not inevitability. It is intervention.

01

Turn U.S. national-defense resilience policy into tested resilience

Fund, deploy and exercise independent power, water, communications, strategic reserves and alternate supply pathways. The Clock moves backward only when mission continuity is demonstrated under realistic disruption.

02

Build whole-of-society outage resilience

Household 72-hour preparedness, community continuity plans and reliable crisis communications can reduce panic, emergency-service overload and demoralization. These measures complement hardened infrastructure; they do not replace it.

03

Use AI for defender scale and deception

Use advanced models for asset discovery, exposure management, monitoring, engineering analysis and restoration planning. Apply CriticalInfra.com’s Burn–Churn–Learn construct to waste adversary compute, create uncertainty and convert deceptive interactions into defensive intelligence.

04

Manage the installed base

Identify provenance, unsupported technology, remote access, embedded components and single points of failure. Prioritize them by mission consequence.

Selected evidence and scenario inputs

Primary sources, assessments and scenario data

CISACross-Sector Cybersecurity Performance GoalsGAOWater & Wastewater Cybersecurity: May 2026FERCCyber and Grid Security / NERC CIPCISA + partnersVolt Typhoon joint advisoryFBI + partnersPLC exploitation across U.S. critical infrastructure: July 2026FBI + EPAWater PLC operational disruptions: July 2026NCSC / AISIFrontier AI cyber capability and ICS evaluation: March 2026AnthropicCountering misuse of AI: September 2026: agentic multi-victim campaignsOpenAIHugging Face incident and the road ahead: August 26, 2026AnthropicGLM-5.3 and the spread of advanced cyber capabilities: September 29, 2026AI 2027Scenario and security forecast: R&D acceleration, model security and diffusion sensitivity analysisDragosAI-assisted ICS targeting: May 2026WaterISACThird-party ICS integrator access and current water-sector OT risk: September 2026METR / RedwoodIndependent OpenAI / Hugging Face incident investigation: August 26, 2026OpenAICoordinated disclosure of critical PHP vulnerabilities: October 2026NSAUrgent advisory on critical PHP vulnerabilities: October 21, 2026NCSC-NLPHP vulnerabilities: exploitation complexity and immediate patching guidance: October 23, 2026SiemensIndustrial product advisory for PHP vulnerabilities: October 23, 2026Schneider ElectricIndustrial cybersecurity notifications: affected product advisoriesAWSPHP2Shell exploitation attempt observed before public disclosure: October 2026AnthropicCoordinated vulnerability disclosure dashboard: 6,157 AI-found vulnerabilities across 591 projects as of October 2, 2026ReutersFrench supermarket refrigeration cyberattack: distributed physical effects, food losses and fires: October 12, 2026AnthropicSelf-replicating AI across commercial IoT devices: experimental research: October 12, 2026MicrosoftStorm-0940 uses ARTEX agentic AI for machine-speed nation-state operations and supply-chain compromise: October 21, 2026MicrosoftLiteLLM 1.107.0 AI infrastructure supply-chain compromise: October 23, 2026Unit 42PHP2Shell threat brief: agentic-AI-style post-exploitation observed after initial compromise: October 23, 2026U.S. ArmyFORTRESS America launch: installation energy independence and hardening: September 30, 2026DefenseScoopFORTRESS America Program Office operationalization: October 7, 2026Fox News / PentagonReported 90-day installation self-sufficiency target and industrial stress tests: October 25, 2026UK Cabinet OfficeHousehold emergency-preparedness campaign for disrupted power, water, phones and internet: October 22, 2026Netherlands GovernmentDenk vooruit campaign: prepare for 72 hours without power, water or internet: October 12, 2026European CommissionPreparedness Union Strategy: minimum 72-hour household supplies and essential-service resilienceGermany BBKUpdated civil-protection guidance: aim for 10 days of self-sufficiency, at least three days as a baselineGoogle Threat IntelligenceARTEX and friends: agentic offensive frameworks observed across real intrusion stages: October 30, 2026UK NCSCArtificial intelligence and cyber intrusion: independent 2028 CNI horizon and AI-cyber drivers: November 3, 2026ReutersEuropean refrigeration campaign reaches French blood-service cold storage: November 4, 2026ReutersFrance attributes refrigeration campaign to IRGC-linked Shahid Kaveh unit: November 9, 2026SANS2026 ICS/OT survey: incident frequency, operational disruption, visibility and detection lag: November 10, 2026Claroty Team82Firmware to Root in 30 Minutes: AI-assisted OT vulnerability research and attack economics: October 13, 2026Booz Allen HamiltonControlled frontier-AI testing in OT/ICS and physical equipment: October 2026Booz Allen HamiltonThe Offensive Frontier: AI as the Attacker: Cyber Weapon Index (CWI), autonomous offense and counter-AI testingAnthropicCyber Verification Program / Project Glasswing: tiered trusted access for verified defenders: October 6, 2026OpenAIDaybreak Blue and Daybreak Red: trusted access for authorized defensive and red-team cybersecurity workBreaking DefenseFORTRESS America public-private financing approach: October 30, 2026U.S. ArmyEnhanced-use lease awards use private capital for installation energy resilience: September 18, 2026CISACI Fortify: isolation, recovery, critical-customer prioritization, manual transition and no-phone/no-internet functional testingCISACyber Resilience Review Service Continuity guide: tabletop, partial-, full- and integrated-recovery exercisesCISA / NSTACIT–OT Convergence Report: resilience under stress and graceful degradationWIREDAbliteration AI, commercial safeguard removal, legitimate-use arguments and operational misuse risk: November 10, 2026a16z SpeedrunSR007 portfolio listing: Abliteration.ai, unrestricted LLMs for cybersecurity, agent testing and defenseAlice ResearchActivation-space abliteration testing: harmful-request compliance increased from 5.8% to 98.0% across five safety-trained modelsWIRED / Tech Against TerrorismAbliterated open-weight variants appearing in under three days on average versus more than 24 days in 2024: October 14, 2026FAANationwide ground stop after communications issues affected multiple Air Route Traffic Control Centers: November 13, 2026ReutersFAA outage investigation: communications-network overload during copper-to-fiber conversion; AI scheduling and cyberattack not blamed: November 14, 2026U.S. GAOFAA modernization: 11,389 project schedules were not integrated; master scheduling needed to manage dependencies and reduce operational disruption: September 2026NATOApproach to counter information threats: monitoring, early warning, coordinated response and resilience against hostile information activityFEMANIMS guidance for public information officers: social listening for rumors, misinformation, information gaps, public perception and critical-infrastructure statusReutersKirsten Davies to serve temporarily as National Cyber Director: November 19, 2026U.S. Department of War CIOKirsten Davies biography: cybersecurity, enterprise technology, legacy modernization and large-scale organizational transformationWhite House ONCDNational Cyber Director role: whole-of-government cyber policy, emerging technology and national cybersecurity coordination