The CISA Act’s cloud clause mattered more than its new name
The November 2018 law renamed a DHS directorate and granted no new powers or money. Its sleeper was a report on how federal defenses would follow agencies into the cloud: who watches, where sensors sit, and what can be seen.
Looking Back revisits past weeks one at a time, up to the most recent, and asks what it meant for the banks, health systems and public agencies we serve. The series was written in 2026, with the benefit of hindsight, and each article says which week it looks back on.
Two-minute summaryThe dek and the first sentence of each section, quoted from the article
The November 2018 law renamed a DHS directorate and granted no new powers or money. Its sleeper was a report on how federal defenses would follow agencies into the cloud: who watches, where sensors sit, and what can be seen.
- The week On November 16, 2018, the Cybersecurity and Infrastructure Security Agency Act of 2018 was approved and became Public Law 115-278.
- A new name on old authorities The act’s rule of construction is unusually direct.
- Who watches: security operations as a service The first item in the required report was a plan for ensuring access to a security operations center as a service.
- Where the sensors sit: intrusion detection after the perimeter The second question in Section 4 was how DHS would adapt its intrusion detection and prevention system to civilian networks in a cloud environment.
- What can be seen: continuous diagnostics in someone else’s data center The third question in Section 4 concerned the Continuous Diagnostics and Mitigation program.
- The case against reading so much into one section There is a reasonable objection to this account.
- With hindsight With hindsight, the act’s structural promises took longer than planned, and its cloud question was answered in stages.
The week
On November 16, 2018, the Cybersecurity and Infrastructure Security Agency Act of 2018 was approved and became Public Law 115-278. It had started in the House as H.R. 3359. The law amended the Homeland Security Act of 2002 by adding a new title for a new agency inside the Department of Homeland Security.
Its first operative step was a renaming. The National Protection and Programs Directorate would, from the date of enactment, be known as the Cybersecurity and Infrastructure Security Agency, or CISA. Any reference to the old directorate in any law, regulation, map, document or record would be read as a reference to the new agency. References to the old Under Secretary were redirected in the same way, to a new Director of Cybersecurity and Infrastructure Security who reports to the Secretary.
The act set out a structure. It created a Deputy Director and named three divisions: Cybersecurity, Infrastructure Security, and Emergency Communications, each headed by an Assistant Director. The heads of the Cybersecurity and Infrastructure Security divisions were to be at the level of an Assistant Secretary, appointed by the President without Senate confirmation. The act also gave the agency its own Privacy Officer. It moved the Office of Biometric Identity Management out of the directorate to the Department’s Management Directorate, and it set a process for deciding where the Federal Protective Service belonged.
The Director’s duties were listed at length. They included leading the agency’s cybersecurity and critical infrastructure programs and coordinating with the Sector-Specific Agencies. They also included carrying out the Secretary’s responsibilities to secure federal information systems under title 44 of the U.S. Code and the Cybersecurity Act of 2015. On request, the agency was to give analysis, expertise and technical assistance to critical infrastructure owners and operators.
Two short sections set the limits. Section 5 said nothing in the act conferred new programmatic, regulatory or enforcement authorities on the Secretary beyond those that existed the day before enactment. Nor was it to affect the existing authority of any other federal agency or DHS component. Section 6 authorized no additional funds. The act was to be carried out with amounts otherwise authorized.
Section 4 drew less attention. Within 120 days, the Secretary had to send House and Senate committees a report on the Department’s leadership role in cloud-based cybersecurity for civilian agencies. The report was to be prepared in coordination with OMB and the General Services Administration. The report had to describe the plan for a security operations center as a service, as called for in the American Technology Council’s December 19, 2017 report to the President on federal IT modernization. It had to name the capabilities DHS would prioritize and the criteria it would use to judge private-sector offerings. It had to explain how government and private capabilities would give consistent visibility across cloud and on-premises environments. And it had to explain how DHS would adapt its intrusion detection and prevention system and its Continuous Diagnostics and Mitigation program to secure civilian networks in a cloud environment.
For most agency technology teams, the week brought a new name on familiar programs. The part of the law that asked them, and DHS, something difficult was the cloud report.
The Cybersecurity and Infrastructure Security Agency Act is best known as the law that created CISA. Read closely, it created very little. It renamed a directorate, rearranged its divisions and stated plainly that no new authorities and no new money came with the change. The section with the most practical consequence for agency technology leaders was a request for a report.
That report asked how the federal government’s shared defenses would follow agencies into commercial cloud services. The question had three parts, and they map onto three separate pieces of an agency’s security architecture. The first is who watches: whether security monitoring would be run by each agency, by DHS, or bought as a service. The second is where the sensors sit: what happens to intrusion detection built for a network perimeter when traffic no longer crosses it. The third is what can be seen: whether continuous diagnostics built to scan agency-owned machines could see assets and settings inside someone else’s data center.
This article takes those three in turn, after a first section on what the act did and did not change. A final section weighs the strongest case against reading so much into one clause. The three questions do not overlap. Monitoring is about people and operations. Sensing is about network paths. Diagnostics are about asset and configuration data. An agency can answer one well and fail the others.
A new name on old authorities
The act’s rule of construction is unusually direct. Nothing in it confers new programmatic, regulatory or enforcement authority on the Secretary of Homeland Security. Nothing in it affects another agency’s existing authority. The funding section adds that no additional appropriations are authorized. For an agency chief information officer, that meant the legal basis of the relationship with DHS did not move in November 2018.
The Director’s duty to secure federal information systems was tied to existing law: the federal information security provisions of title 44 and the Cybersecurity Act of 2015. The binding operational directive, the instrument through which DHS directs agencies’ security actions, already came from title 44. CISA’s own description of a later directive cites the Secretary’s authority under 44 U.S.C. 3553(b)(2) to develop and oversee such directives, and the agencies’ duty to comply under section 3554. Those provisions predate the CISA Act. The renaming changed the letterhead on directives, not their force.
What changed for agencies
Three things did change, and each mattered at the margin. The first was visibility. A named agency with a Director is easier for agency leaders, states and companies to find and cite than a directorate inside a department. The second was internal structure. Separate divisions for cybersecurity and infrastructure security gave each a statutory head, which affects who signs and who answers for programs. The third was focus. Moving biometric identity management out, and setting a process for the Federal Protective Service, narrowed the agency toward its cyber and infrastructure mission.
None of those changes altered what an agency had to do on Monday morning. FISMA reporting continued. Directives continued. The Continuous Diagnostics and Mitigation program, which CISA says was developed in 2012, continued. Intrusion detection at agency internet gateways continued. An agency compliance officer reading the act would have found no new obligation to map.
Why the absence of new powers mattered
The lack of new authority shaped the cloud question in a specific way. DHS’s tools for civilian agencies were built on an arrangement in which agency traffic passed through a limited number of managed gateways. The Trusted Internet Connections initiative, CISA says, has since 2007 consolidated network connections and enhanced visibility across the federal network. Sensors at those gateways could see agency traffic because the traffic had to pass them.
Cloud services broke that arrangement without any change in law. When an agency moved email or collaboration to a commercial provider, much of the traffic went from users to the provider directly. The legal authority to direct agencies stayed the same. The physical ability to see what happened did not. The CISA Act gave DHS no new power to require providers or agencies to route traffic differently. Any answer had to come from policy, procurement and architecture, which is why Section 4 asked for a plan rather than conferring a mandate.
Coordination without command
The act’s savings clause reinforced the point. Nothing in the new title was to affect the existing authority of any other component of the Department or any other federal agency. The Director’s duties toward Sector-Specific Agencies were framed as coordination: developing mechanisms for frequent collaboration, coordinating outreach, and providing assistance in coordination with them where appropriate.
For a bank, a hospital or a utility, this meant the agency it answered to did not change. Its regulators kept their authority. What changed was that a named federal agency now had a statutory duty to coordinate, share analysis and offer help. That is useful, but it is help on request, not supervision. A private operator’s security program still had to satisfy its own regulator first. The cloud question that the act raised for federal agencies reached private operators later, and through other channels, such as their regulators’ guidance and their providers’ defaults.
The difference for states, localities and contractors
The act’s limits also explain why its effect differed by type of organization. For federal civilian agencies, CISA had directive authority through title 44, and the cloud question was whether that authority could be exercised over cloud tenants. Later directives answered by naming cloud tenants explicitly. For state and local governments, the nearest duty in the act was narrower: providing, on request, analyses, expertise and technical assistance to critical infrastructure owners and operators. A county or a state agency that operates critical infrastructure could ask for help, but nothing in the act required it to accept any CISA tool or setting.
Directives also have stated limits within the federal government. CISA’s text of a 2024 directive notes that directives do not apply to statutorily defined national security systems or to certain Defense Department and intelligence community systems. A contractor running a federal system inherits obligations through its contract and the agency’s authorization, not from the CISA Act. That leaves three different routes by which the same cloud configuration question reaches an organization: a directive, a request for assistance, or a contract clause. Organizations should check with counsel which route applies to them.
Who watches: security operations as a service
The first item in the required report was a plan for ensuring access to a security operations center as a service. The act tied the phrase to the American Technology Council’s report on federal IT modernization, which it cites by date. The idea, as we read it, was that a small agency should not need to build its own round-the-clock monitoring team to get round-the-clock monitoring.
The question behind it is about people and contracts, not technology. A security operations center needs analysts on shift, playbooks, authority to act, and access to the logs that matter. A large department may be able to staff that. A small board or commission with a handful of technical staff may not, and moving its data into cloud services whose logs it has never collected makes the gap wider.
Three ways to buy watching
An agency deciding who watches its cloud services has three options, and each moves accountability differently. It can build an internal team, which keeps control and context but costs the most and is hardest to staff at night. It can buy a managed service from a commercial provider, which buys coverage but moves the analyst away from the agency’s own knowledge of its systems. Or it can use a shared federal service, which promises consistency but depends on the provider agency’s priorities and funding.
An agency that mixes the options creates seams. The internal team owns the on-premises network. A vendor watches endpoints. A shared service receives some logs. Each party sees part of an incident. The practical test is whether anyone can see an intrusion that starts with a stolen cloud credential, moves through email and ends in a file share. If each of those systems is watched by a different party, the answer may be no.
What the contract has to say
Whichever option an agency chooses, the contract or agreement carries the weight. It has to say which logs the watcher receives and how quickly. It has to say what the watcher may do on its own authority, such as disabling an account, and what needs agency approval. It has to set how quickly the agency hears about a confirmed incident and who decides when to tell the agency’s leadership and DHS.
It also has to say what happens to the data. Logs from a cloud service contain user names, message metadata and sometimes content. An agency that sends them to a third party needs to know where they are stored, who can read them and how long they are kept. Those are privacy questions as much as security ones. The CISA Act itself gave the agency a Privacy Officer charged with assuring that its technologies sustain privacy protections, a reminder that monitoring programs carry that duty too.
How small agencies differ from large ones
The difference by agency size is large. A cabinet department may run several operations centers and want a shared service only to fill gaps. A small agency may have one security officer who also runs the help desk. For that agency, the useful service is complete: monitoring, triage, and an analyst who calls the agency when something matters. A tool without the analyst moves the problem rather than solving it.
CISA’s Continuous Diagnostics and Mitigation program later described a Shared Services Platform giving 75 eligible non-CFO Act agencies access to its capabilities. This arrangement shows the government accepted the small-agency problem as structural. It does not, by itself, show that every small agency gained full monitoring of its cloud services. An agency leader should judge the arrangement by the test above: whether a cloud-originated intrusion would be seen, by whom, and how fast.
The dependency a shared service creates
Shared monitoring also creates a dependency that, in our view, deserves more weight than it gets. If a central service sets the detection rules, the agency inherits its priorities. A rule tuned for the threats most common across government may miss what is specific to one agency’s mission. A grants system may attract fraud; a public portal may attract credential stuffing.
The remedy is a division of labor written down in advance. The shared service runs common detections and keeps the shift coverage. The agency supplies context: which accounts are privileged, which systems are critical, which data is sensitive, and what normal looks like for its users. It also keeps the authority to add its own detections. Without that context, a shared analyst sees an unusual login and cannot tell whether it is a traveling executive or an intruder.
Funding is the other dependency. The act authorized no new money, so shared services had to be paid for from existing appropriations, by the provider or by the agencies that used them. A small agency that relies on a shared service should ask what happens if the provider’s budget shrinks. The answer belongs in its continuity planning, alongside the more familiar question of what happens if a vendor fails.
The metric that matters
The most useful single measure for the watching question is coverage of critical cloud services by an accountable monitor. It can be tracked as a list: each cloud tenant that holds sensitive data, the logs it produces, where those logs go, and the named party that reviews them. Gaps on that list are the honest answer to whether the agency has a security operations center for its cloud estate, whatever the contract is called.
Where the sensors sit: intrusion detection after the perimeter
The second question in Section 4 was how DHS would adapt its intrusion detection and prevention system to civilian networks in a cloud environment. That system was built around gateways. OMB’s later memorandum M-19-26, issued September 12, 2019, described the earlier policy plainly. Previous memoranda required agency traffic to flow through a physical TIC access point, which had proven to be an obstacle to cloud adoption.
The obstacle was a matter of physics and cost. Routing a remote user’s traffic to a government gateway and back out to a commercial cloud service adds distance, delay and bandwidth charges. It also makes the gateway a single point of failure for services designed to be reached from anywhere. Agencies that complied fully paid in performance. Agencies that made exceptions paid in visibility.
From one perimeter to several use cases
M-19-26 rescinded four earlier TIC memoranda and replaced the single model with use cases. DHS would define the use cases. They would set out which alternative security controls, such as endpoint and user-based protections, had to be in place where traffic did not pass through a physical TIC access point. The initial list named a traditional default, cloud use cases for infrastructure, platform, software and email as a service, an agency branch office case, and a remote user case.
The memorandum also kept a link to the government-wide sensors. TIC use cases would highlight scenarios where agencies met government-wide intrusion detection and prevention requirements without routing traffic through a gateway service. The memorandum named the National Cybersecurity Protection System, including EINSTEIN, as an example. The cloud report’s second question had its answer in policy: sensing could move, but it could not disappear.
What moving the sensor means operationally
For an agency, moving from a gateway to use cases changes who operates the control. At a gateway, a central provider ran the sensor and the agency mostly consumed alerts. Under a cloud use case, some of the equivalent protection comes from the agency’s own endpoint agents, identity controls and the cloud provider’s logging. The agency becomes responsible for assembling those pieces and proving they work together.
That proof has a specific form. M-19-26 told agency chief information officers to keep an accurate inventory of network connections, with the service provider, cost, capacity, traffic volume, configurations and topology of each. It told agencies to update their network and system boundary policies within one year and to identify which use cases they would allow. OMB and DHS would track implementation through FISMA reporting. The inventory is the evidence an auditor will ask for first, because a sensor cannot cover a connection nobody listed.
The trade-off the use cases created
The use cases solved the performance problem and created a consistency problem. One perimeter meant one set of controls that everyone could inspect. Several use cases meant several combinations of controls, each implemented differently by each agency. Two agencies could both claim compliance with the cloud use case while seeing very different amounts of their traffic.
M-19-26 anticipated part of this by requiring DHS to build a compliance verification process for each use case and by asking for continuous validation rather than point-in-time checks. The burden of that validation falls on agency teams. They have to show, for each path from user to service, which control sees the traffic and where its records go. Showing it is harder than pointing at a gateway, and it is the price of letting traffic take the short route.
Encrypted traffic and the inspection question
Gateways had another quality that use cases lose. A gateway could, in principle, inspect traffic in transit. Where traffic to a commercial cloud service is encrypted between the user and the service, an agency that sends users directly to the service gives up the chance to inspect that traffic on the way. What remains is what the endpoint sees, what the identity system records and what the provider logs.
That changes which logs matter. Under a gateway model, network records were the primary evidence. Under use cases, the identity provider’s sign-in records, the endpoint agent’s process records and the cloud service’s audit log carry most of the weight. An agency that has not arranged to keep all three, for long enough and in one place, has moved its traffic without moving its evidence. Retention periods, export formats and who pays for storage become architecture decisions, not housekeeping.
Branch offices and remote users
The two non-cloud use cases in M-19-26 point to the same shift. The branch office case assumed an office that uses headquarters for most services and supports software-defined wide area network technologies. The remote user case grew out of earlier FedRAMP TIC overlay work. It showed how a user on government furnished equipment connects to the agency network, cloud services and the internet.
Both cases move protection toward the device and the user. The device has to be managed, patched and monitored wherever it is. The user has to authenticate strongly wherever they are. For an agency with field offices, inspectors or caseworkers, that means the endpoint program is now part of the network security program. Where budgets and teams still treat them separately, the 2019 policy exposes a structural gap.
How a state agency or county reads this
TIC policy applies to federal agencies. State and local governments were never bound by it. The architecture question is the same for them, though, and arguably sharper, because no federal gateway mandate ever applied to them. A county moving its email to a commercial service faces the same choice between routing for visibility and going direct for performance. The federal use cases offer a published model a county can borrow, with no obligation to adopt it.
What can be seen: continuous diagnostics in someone else’s data center
The third question in Section 4 concerned the Continuous Diagnostics and Mitigation program. CISA describes CDM as delivering tools, integration services and dashboards that help agencies reduce their threat surface, increase visibility into the federal security posture, improve response and streamline FISMA reporting. Its published capability areas are asset management, identity and access management, network security management and data protection management.
Diagnostics in the original sense meant finding the devices on an agency network, checking their software and settings, and reporting the results to agency and federal dashboards. That works when the agency owns the devices and can install agents or scan them. Cloud services complicate each step. In software as a service, there is no server for the agency to scan. What the agency controls is a tenant: its users, its sharing settings, its logging choices and its security policies.
From assets to configurations
The shift that followed was from counting machines to checking settings. In a commercial collaboration suite, the risks that matter are configuration choices. Whether legacy sign-in methods are allowed, whether external sharing is open, whether administrative roles require stronger authentication, and whether audit logs are retained are all settings. They can drift with every vendor update and every administrator’s change.
CISA’s Secure Cloud Business Applications project, which the agency says it established in 2022, addressed exactly this. It published secure configuration baselines for Microsoft 365 and Google Workspace. In 2023 it added two assessment tools, ScubaGear for Microsoft 365 tenants and ScubaGoggles for Google Workspace tenants. CISA says the project responded to security and visibility gaps exposed by intrusions into software-as-a-service environments.
When the baseline became binding
On December 17, 2024, CISA issued Binding Operational Directive 25-01, Implementing Secure Practices for Cloud Services. It requires federal civilian agencies to implement the mandatory policies in the SCuBA baselines, marked as “shall” actions, for in-scope cloud tenants. At issuance, the final baselines in scope were those for Microsoft Office 365. Agencies had to identify their in-scope tenants by February 21, 2025, and deploy the assessment tools by April 25, 2025.
The directive also required agencies to begin continuous reporting, either by integrating tool results with CISA’s continuous monitoring solution or by reporting results manually, and to remediate deviations. It states that it complements FedRAMP, NIST guidance and the TIC 3.0 cloud use case. Read alongside the 2018 act, the directive answers Section 4’s third question. Diagnostics in the cloud would mean measuring tenant settings against a published baseline and sending the results to a shared view.
What an agency has to run to comply
The operational work behind that answer has four parts. The agency needs a complete list of its cloud tenants, which is harder than it sounds where components bought services separately. It needs the assessment tool running against each tenant on a schedule. It needs a path for the results to reach CISA’s monitoring. And it needs owners who will fix deviations, which can mean changing settings that users have come to rely on.
That last part is where the cost falls. Turning off a legacy sign-in method can break an old scanner or a shared mailbox script. Restricting external sharing can disrupt a program office that works with grantees. Each fix needs a business owner’s agreement and sometimes a replacement process. An agency that treats the baseline as a security team’s project will find its deviations stay open. One that assigns each deviation to the owner of the affected business process can close them.
Evidence an inspector general will ask for
For audit purposes, the cloud diagnostics question produces a short evidence list. It includes the tenant inventory with owners, the latest assessment results for each tenant, the record of deviations and their remediation dates, and any documented exceptions with the reason and the approving official. It also includes proof that results reach the shared monitoring view. Those records answer the question Section 4 posed in 2018: whether DHS, and the agency, can see what is happening in the cloud.
Exceptions deserve their own scrutiny. Sometimes a baseline policy cannot be met because a business process depends on an older method. That exception is legitimate only if it is documented, approved at the right level, time-limited and paired with a compensating control. An exception register that grows every quarter and never shrinks is a sign that the agency has accepted a risk without deciding to.
Baselines that expire
BOD 25-01 contains a detail with large operational consequences. Any baseline not updated within one year automatically falls out of scope and is removed from the catalog of required configurations. CISA may also publish new baselines, and once it does, the products they cover come into scope.
For agencies, that makes the directive a moving target by design. A tenant that passed last year may fail this year because the baseline changed, or a product the agency treated as out of scope may come in. The compliance program has to watch the catalog as well as the tenants. The right owner for that watch is whoever already tracks directives and FISMA metrics, with a standing link to the cloud administrators who make the changes.
Several tenants, several owners
A large department may hold several tenants of the same service, bought by different components at different times. Each may have its own administrators, its own exceptions and its own history. The directive’s first required action, a list of tenants with the owning agency or component for each, exposes that history. For some departments the list is the most useful product of the whole exercise, because it shows where consolidation would reduce both cost and risk.
Consolidation has its own risks. Merging tenants moves mailboxes, files and identities, and a poorly sequenced migration can create the very misconfigurations the baseline is meant to prevent. A department should finish measuring each tenant against the baseline before merging any, so that it knows which settings it is carrying forward and which it is fixing.
The limit of a baseline
A configuration baseline sees settings, not behavior. It can confirm that strong authentication is required for administrators. It cannot confirm that an administrator’s session has not been hijacked. For this reason the third question depends on the first two. Diagnostics tell an agency its tenant is configured as intended. Sensors and watchers tell it whether someone is abusing that configuration. A program that funds only one of the three leaves a predictable gap.
The case against reading so much into one section
There is a reasonable objection to this account. Section 4 required only a report to congressional committees. The policies that followed, from M-19-26 to the cloud reference architecture to BOD 25-01, had their own drivers. Executive Order 14028 is the basis CISA cites for its Cloud Security Technical Reference Architecture. Intrusions into cloud services, not a 2018 report, are the reason CISA gives for the SCuBA project. On this view, the act merely asked a question that events would have forced anyway.
The objection is partly right. A report request does not cause an architecture. The 2018 clause is better read as a marker of where the problem stood. By November 2018, Congress had written into law that DHS, working with OMB and GSA, must explain how federal defenses built for a perimeter would work in the cloud, and the act wrote down the three components that answer would need. What followed answered those components in that order: policy for routing and sensing, then a reference architecture, then binding configuration rules.
A second objection: the reorganization was the main event
A second objection holds that the reorganization itself was the act’s major contribution, and that the cloud clause is a footnote. GAO’s March 2021 review supports part of that view. It found that CISA had launched a three-phase transformation to implement the act, had consolidated multiple incident response centers and consolidated points of contact for infrastructure security stakeholders. It also found the transformation incomplete, with 37 of 94 planned phase-three tasks done by mid-February 2021 and 42 past their planned dates.
Those findings cut both ways. A reorganization that took years to complete, and that GAO tied to the agency’s ability to respond to incidents, did matter. But it mattered mostly inside CISA. For an agency CIO, a state security officer or a contractor, the external effect of the act came through the programs the agency ran. The cloud question shaped those programs more than the organization chart did.
A third objection: cloud providers solved this themselves
A third objection is that commercial cloud providers built strong native security tools, so federal sensors and baselines were redundant. Provider tools are real and useful. They do not settle who watches, who sets the baseline, or who can compare one agency’s posture with another’s. A provider’s tools report to its customer. A government-wide view requires that results flow to a common place under common rules. That is the role the 2018 clause anticipated for DHS, and it remains a role no single provider can play.
What we would tell an agency CIO today
The practical reading of the act is that the three questions never went away, and that each needs a named owner. Watching belongs to whoever runs security operations, whether staff, vendor or shared service, and the evidence is a coverage list. Sensing belongs to the network and endpoint teams together, and the evidence is the connection inventory M-19-26 asked for. Seeing belongs to the cloud administrators, and the evidence is the baseline assessment record.
An agency that can produce all three records quickly is in a strong position for an inspector general, a CISA review or an incident. An agency that cannot produce one of them has found its gap. The order in which to close gaps follows risk: the tenants with the most sensitive data first, the paths used by privileged accounts next, and the long tail of low-risk services last.
What would change the argument
The argument would be weaker if agencies had kept routing cloud traffic through gateways at scale, or if diagnostics had remained limited to agency-owned machines. Published federal policy went the other way, toward use cases, a reference architecture and binding cloud configuration baselines. Each step answered one of the questions the act asked in its least-quoted section.
With hindsight
With hindsight, the act’s structural promises took longer than planned, and its cloud question was answered in stages. GAO reported on March 10, 2021 that CISA had intended to complete its organizational transformation by December 2020. By mid-February 2021, it had finished 37 of 94 planned tasks for the final phase. GAO said that until CISA set updated milestones and an overall deadline, it would be hindered in meeting the goals of its transformation. It linked this to the agency’s ability to respond to incidents such as the cyberattack discovered in December 2020. GAO also found workforce planning only partly addressed, a notable gap for an agency that depends on scarce experts. On the positive side, GAO recorded that the first two phases had produced a new organization chart, consolidated several incident response centers and given infrastructure security stakeholders fewer points of contact.
The routing and sensing question was answered first. On September 12, 2019, OMB issued M-19-26, which rescinded the earlier TIC memoranda that had required traffic to flow through a physical access point. It replaced them with use cases, including cloud, branch office and remote user cases, and required agencies to keep an inventory of network connections. CISA describes TIC 3.0 as highly iterative, with guidance that reflects new processes and technology as they become available, and its current TIC 3.0 training includes a Cloud Use Case.
The architecture question came next. CISA published version 2 of its Cloud Security Technical Reference Architecture, with a revision date of June 21, 2022, under Executive Order 14028, to guide agencies in adopting cloud technology and moving toward zero trust. CISA’s website now marks that document as archived content. On the same agency’s website, the TIC 3.0 core guidance documents were updated to reflect organizational changes at CISA, changes CISA described as primarily administrative, with no technical content changes.
The diagnostics question was answered last and most firmly. CISA’s SCuBA project, started in 2022, produced secure configuration baselines and, in 2023, assessment tools. Binding Operational Directive 25-01, issued December 17, 2024, made the mandatory baseline policies compulsory for in-scope federal civilian cloud tenants. Tenant inventories were due February 21, 2025, and assessment tools by April 25, 2025. As a binding operational directive, it is a compulsory direction to federal executive branch agencies, and it does not apply to statutorily defined national security systems. The directive also builds in its own expiry: baselines not updated within a year drop out of scope. CISA says that although SCuBA’s primary goal is to secure federal civilian information, any organization can use it, which gives states, localities and private operators a published benchmark without an obligation.
What held up was the act’s framing: that federal cyber defense in the cloud had to answer who watches, where sensing happens and what can be seen. What did not hold up was the hope that a renaming without new money would be quick. The CDM program continued, and CISA describes a Shared Services Platform that gives 75 eligible non-CFO Act agencies access to its capabilities. Whether every agency’s cloud estate is fully watched is not something public sources let us confirm. The evidence shows a steady move from gateways to settings, with agencies carrying more of the operational work than in 2018.
Spotted an error? Tell us and we will check it and correct the article where needed.