An August 2017 Senate bill put connected-equipment security in a vendor’s signed promise to federal buyers
On August 1, 2017, a Senate bill proposed four promises vendors would certify for connected equipment sold to agencies. In our reading, the enacted 2020 law moved that test to agency CIOs, and both versions depended on inventories many agencies had not completed by 2026.
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
On August 1, 2017, a Senate bill proposed four promises vendors would certify for connected equipment sold to agencies. In our reading, the enacted 2020 law moved that test to agency CIOs, and both versions depended on inventories many agencies had not completed by 2026.
- The week On August 1, 2017, four senators introduced S. 1691, the Internet of Things (IoT) Cybersecurity Improvement Act of 2017.
- The baseline: four promises at the time of the proposal The bill’s central clause asked the contractor to certify four things in writing about the device. [1] Each addressed a different architectural weakness.
- The lifecycle: duties after the sale The second part concerns what happens once the device is installed.
- The exceptions: buying what does not meet the baseline The third part concerns deviation.
- The visibility: inventories, researchers and the database The fourth part concerns information.
- How the approach reached different organizations The bill was about federal purchasing, but its effects would reach well past federal buyers.
- Objections, in order Our reading has reasonable critics.
- What an agency architect could check The bill’s four parts translate into questions any organization can ask about its connected devices, whether or not a law requires it.
- With hindsight The law that was enacted came from a later bill.
The week
On August 1, 2017, four senators introduced S. 1691, the Internet of Things (IoT) Cybersecurity Improvement Act of 2017. The bill was referred to the Senate Committee on Homeland Security and Governmental Affairs. [1] Its long title described its aim: minimal cybersecurity operational standards for connected equipment that federal agencies buy. [1]
The bill did not regulate manufacturers directly. It worked through federal contracts. It directed the Director of the Office of Management and Budget to issue guidelines, within 180 days of enactment, requiring each executive agency to put specific clauses into such contracts. [1] The budget office was to consult the Secretary of Defense, the Administrator of General Services, the Secretary of Commerce, the Secretary of Homeland Security and any other intelligence or national security agency it considered necessary. [1]
The bill defined an internet-connected device as a physical object capable of connecting to, and in regular connection with, the internet. The object also had to have processing capabilities that can collect, send or receive data. [1] It also defined firmware, hardware, software, a security vulnerability, a fixed or hard-coded credential and a properly authenticated update. [1]
The first required clause was a written certification from the contractor. The product could contain no hardware, software or firmware component with a known flaw listed in the National Vulnerability Database or a similar credible database. [1] It had to accept properly authenticated and trusted updates from the vendor. [1]
The certification also covered protocols and passwords. The product had to use only non-deprecated industry-standard protocols and technologies for communications, encryption and interconnection. It could not include fixed or hard-coded credentials used for remote administration, delivery of updates or communication. [1]
Further clauses covered the device’s life after purchase. The contractor would have to tell the agency of flaws later disclosed to it and keep the software or firmware updatable so flaws could be patched securely. It would have to repair or replace the unit when an update could not fix a flaw. [1] It would also have to say how the unit received updates, when security support was expected to end, and when it had ended. [1]
The bill allowed exceptions. An agency could petition the budget office for a waiver to buy equipment with limited processing and software functionality, under conditions to be defined with the standards agency. The bill listed possible conditions such as network segmentation, system-level controls, multi-factor authentication and gateways that can isolate or disable equipment. [1]
A contractor could also show compliance with an existing third-party security standard of equal or greater strength, with accreditation standards for certifiers set by the standards agency. An agency could keep its own evaluation process if approved. [1] A contractor could disclose a known flaw by applying for a waiver, which the purchasing agency’s head could approve in writing. [1]
The bill reached past purchasing in three ways. It directed the Department of Homeland Security’s National Protection and Programs Directorate to issue guidelines on coordinated vulnerability disclosure for contractors, based in part on the international standard ISO 29147. [1] It amended the Computer Fraud and Abuse Act and the Digital Millennium Copyright Act to shield good-faith research on products of the class, model or type sold to the government. The shield applied if the researcher followed those guidelines. [1]
And it required each agency to keep an inventory of connected equipment bought under the Act, using guidance based on DHS’s Continuous Diagnostics and Mitigation program. The budget office would also keep public databases of products covered by the research protections and of products whose security support had ceased. [1]
A security requirement for a device can live in one of two places. It can be a promise from the seller, written into the contract and certified at the time of purchase. Or it can be a standard the buyer applies, checked by the buyer’s own officials before and after the purchase. The 2017 bill chose the first. The law that passed in 2020 chose the second. [1] [2]
Our argument is that the 2017 bill placed the device security test in the vendor’s signed certification. The enacted law later moved that test to NIST standards and the agency’s chief information officer. The certification’s architectural content still describes what agencies need. [1] [2] We take the bill in four parts:
- The baseline: what a device had to be on the day it was offered.
- The lifecycle: what the vendor had to do after the sale.
- The exceptions: how an agency could buy a device that did not meet the baseline.
- The visibility: how the government would see its devices and learn of their flaws.
The first part covers the device at purchase. The second covers its supported life. The third covers deviations from the rule. The fourth covers the information the government would need to run the rule. Together they span a device from proposal to retirement. After the four parts, we look at how the approach reached different organizations, answer objections, and list what an agency architect could check.
A note on sources and our reading. We describe the 2017 bill from its introduced text. We describe what followed from the 2020 law, NIST’s guidance, CISA’s directive and GAO’s reports. Statements about what the bill or the law would require of a particular agency or contractor are our reading. Agencies and contractors should confirm their obligations with counsel.
The baseline: four promises at the time of the proposal
The bill’s central clause asked the contractor to certify four things in writing about the device. [1] Each addressed a different architectural weakness. Together they described a minimum for any connected device on a federal network.
Why devices differ from servers
The bill’s definitions explain why connected devices needed their own rule. It defined firmware as a program and data stored in hardware, typically in read-only or programmable read-only memory, such that they cannot be dynamically written or modified during execution. [1] Software, by contrast, may be dynamically written or modified. [1]
That distinction is the architecture problem in miniature. A server’s software can be patched by its owner. A device’s firmware may change only through a vendor-supplied image, if at all. The bill’s definition of a hard-coded credential made the same point: a value incapable of being changed by the user except through a firmware update. [1]
In our reading, the definitions show the drafters understood that the owner of a connected device may be unable to fix it. The baseline promises were aimed at the design choices that decide whether the owner, the vendor or no one can respond to a flaw.
No known vulnerabilities
The first promise was that, when the proposal was submitted, the device contained no component with known vulnerabilities or defects. The test was a listing in the National Vulnerability Database or a similar database OMB selected. [1]
This tied a procurement document to a living public database. The certification was dated to the proposal. A vulnerability published the next day would not make it false. In our reading, that made the promise a snapshot, useful for keeping obviously flawed products out, and dependent on the later clauses for anything after.
It also put a demand on the supply chain beneath the vendor. A device is built from components, including third-party chips, libraries and operating systems. To certify no known vulnerabilities in any component, a vendor would need to know what components it shipped. The bill did not use the term, but the certification implied a bill of materials the vendor could check against the database. [1]
Authenticated updates
The second promise was that the device relied on software or firmware capable of accepting properly authenticated and trusted updates from the vendor. [1] The bill defined a properly authenticated update as one with some method of authenticity protection, such as a digital signature, so that unauthorized updates can be automatically detected and rejected. [1]
This is an architectural requirement, not a configuration setting. A device either has a secure update mechanism in its design or it does not. The clause recognized that a device which cannot be patched safely cannot meet any later promise to fix flaws. [1]
Current protocols
The third promise was that the device used only non-deprecated industry-standard protocols and technologies for communications, including standard ports, for encryption and for interconnection with other devices. [1]
In our reading, this clause targeted the long tail of connected equipment that ships with old protocols because they are cheap or familiar. It left open who decides when a protocol is deprecated. The bill did not name an authority, which would have left that judgment to OMB’s guidelines and to individual contracts. [1]
No hard-coded credentials
The fourth promise was that the device included no fixed or hard-coded credentials used for remote administration, the delivery of updates or communication. [1] The bill defined such a credential as a password, token, key or similar value set by the vendor and incapable of being changed or revoked by the user except through a firmware update. [1]
This promise is the narrowest of the four. It names a design flaw by its mechanism and bans it. A device with a vendor-set administrator password that its owner cannot change gives every attacker who learns the password access to every unit. The clause removed that design from federal purchasing, in the bill’s terms. [1]
What certifying would demand of a vendor
Each promise implied work inside the vendor. To certify no known vulnerabilities, the vendor had to match its components against the database at the time of the proposal. To certify authenticated updates, it had to have a signing process and protect the signing keys. To certify no hard-coded credentials, it had to review its firmware for them. [1]
In our reading, the bill would have pushed vendors toward practices that some already followed and others did not: component inventories, signed firmware and credential reviews. A federal certification would have made those practices a condition of the sale, and given a reason to document them.
The timing created its own question. A vulnerability published after the proposal but before the award, or before delivery, would fall outside the certification’s moment. The bill’s notification clause would catch it only once the contract existed. An agency would want its contracting process to recheck the database at award. [1]
Why the baseline was a promise
The baseline took the form of a written certification by the contractor. [1] In our reading, that choice placed the evidence with the seller. The agency would hold a signed statement. It would not, under this clause alone, test the device itself.
A certification has strengths. It is cheap for the buyer, it creates a clear record and it gives the government a contractual remedy if the statement proves false. It also has limits. Its accuracy depends on the vendor’s knowledge of its own components, and its enforcement depends on someone checking it after the fact. [1]
What the 2020 law did with the baseline
The enacted law did not list the four promises. It directed NIST to publish standards and guidelines on agencies’ use and management of IoT devices, including minimum information security requirements, consistent with NIST’s work on secure development, identity management, patching and configuration management. [2]
NIST’s SP 800-213, published in November 2021, frames device security as requirements an organization identifies for a device and its manufacturer, in the context of system risk management. [4] In our reading, the substance of the 2017 baseline survives inside that approach. What changed was who decides whether a device meets it.
The lifecycle: duties after the sale
The second part concerns what happens once the device is installed. Connected devices can stay in service for years. The bill’s further clauses tried to keep the vendor engaged for that period. [1]
Telling the agency about new flaws
One clause required the contractor to notify the purchasing agency of vulnerabilities or defects later disclosed to the vendor by a researcher, or otherwise known to it. The duty lasted for the duration of the contract. [1]
The duty ran for the contract’s term, not the device’s life. In our reading, that gap matters. A device bought under a short contract could remain on the network long after the notification duty ended. An agency would need its own way to learn of flaws in equipment whose contracts had expired.
Keeping the device patchable
Another clause required the device’s software or firmware to be updated or replaced, consistent with the contract’s terms on support, so that any future vulnerability could be patched in a properly authenticated and secure manner. [1] The baseline promised that updates were possible. This clause promised they would be delivered. [1]
Repair or replace
Where a new vulnerability could not be fixed by an update, the contractor had to provide a repair or replacement in a timely manner. [1] The bill did not define timely. In our reading, that word would have carried the weight of the clause in any dispute, and OMB’s guidelines or the contract would have had to fill it in. [1]
This clause addressed a real architectural problem. Some flaws sit in hardware or in code that cannot be updated in the field. For those, the only fix is physical. A contract that requires replacement in that case moves the cost of a design flaw from the buyer to the seller. [1]
The end of support
The final lifecycle clause required general information on the device’s ability to be updated. It covered how the device received security updates, the expected timeline for ending security support, and formal notice when support had ceased. [1] It also allowed for other information recommended by the National Telecommunications and Information Administration. [1]
Coupled with the public database of devices whose support had ceased, this clause gave agencies a date to plan around. [1] In our reading, it was the clause with the clearest architectural value. An agency that knows when a device will stop receiving fixes can plan its replacement, segment it in the meantime, or decline to buy it.
Who pays for years of support
Support has a cost. A vendor that promises updates, notifications and timely replacement prices that promise into the contract. The bill tied the update duty to the contract’s terms on support. [1] In our reading, that meant the length and price of support would become negotiated terms in every device purchase.
For an agency, that is a budgeting change. A cheaper device with a short support period may cost more over its life than a dearer one with longer support, once replacement and isolation costs are counted. The end-of-support disclosure would have given buyers the data to make that comparison. [1]
What the 2020 law did with the lifecycle
The enacted law did not impose lifecycle clauses directly. Its lifecycle mechanism was the disclosure strand: NIST guidelines for contractors and subcontractors on receiving and disseminating information about vulnerabilities and their resolution, overseen by OMB and administered by DHS. [2]
It also required the Federal Acquisition Regulation to be revised as necessary to implement those guidelines. [2] In our reading, the 2017 bill’s notification clause became a disclosure process, applied through standards and acquisition rules rather than written device by device into each contract.
The exceptions: buying what does not meet the baseline
The third part concerns deviation. Not every device can meet every promise. The bill offered several routes around the baseline, each with its own conditions. [1]
Devices too simple to comply
An agency that reasonably believed a compliant device with limited data processing and software functionality would be unfeasible or economically impractical could petition OMB for a waiver to buy a non-compliant device. [1]
OMB, in close coordination with NIST, was to define conditions under which such a device could be used with security equivalent to the baseline. The bill listed conditions OMB might consider: network segmentation or micro-segmentation, system-level controls including operating system containers and microservices, multi-factor authentication, and gateways and edge systems that can isolate, disable or remediate devices. [1]
In our reading, this provision was architectural at heart. It accepted that some devices would be weak, and moved the protection to the network around them. A device that cannot defend itself can be placed where it cannot reach anything that matters. The bill also allowed OMB to set deadlines for removing or disconnecting non-compliant devices. [1]
Compensating controls, illustrated
Consider a hypothetical building controller that cannot accept signed updates. Under the bill’s exception, an agency might place it on its own network segment, reachable only through a gateway that logs and filters traffic, with administrative access requiring multi-factor authentication. [1]
Each of those controls appears on the bill’s list of conditions OMB might adopt. [1] Together they convert a device-level weakness into a network-level design. The cost moves from the vendor to the agency’s architecture team, which must build and maintain the segment and the gateway.
In our reading, the exception also creates a long-term commitment. A segmented weak device must stay segmented for its life, through network changes, reorganizations and staff turnover. The bill’s option of deadlines for removing non-compliant devices recognized that such commitments erode. [1]
Equivalent standards
A contractor could instead show compliance with an existing third-party security standard providing equal or greater security, with third-party written certification. NIST, with OMB and other agencies, would set accreditation standards for certifiers and decide whether the standards were aligned with the guidelines. [1]
An agency with its own evaluation process for connected devices could keep using it, with OMB approval, if NIST found it aligned. [1] These routes recognized existing programs. They also multiplied the number of acceptable paths, each needing its own oversight.
Disclosing a known flaw
The bill created a route for honesty. A contractor could apply for a waiver of the no-known-vulnerabilities promise, for the purpose of disclosing a known vulnerability. The application had to identify the vulnerability, describe mitigations and justify secure use of the device despite it. [1]
If the purchasing agency’s head approved, the agency would give the contractor a written statement accepting the risk as represented. [1] In our reading, this provision is underrated. It turned a vendor’s knowledge of a flaw from a liability into a disclosure, and put the risk decision with a named official.
Waivers as a later lever
The bill also gave OMB, beginning five years after enactment, authority to waive the guidelines in whole or in part for an agency, and required a report to Congress on their effectiveness by that time. [1] The design expected the rule to be revisited. It did not expect it to stay fixed.
What the 2020 law did with exceptions
The enacted law kept a waiver, but placed it with the agency. The agency head could waive the procurement prohibition if the CIO determined the waiver was needed for national security or research. A waiver was also possible if the CIO determined the device was secured using alternative and effective methods appropriate to its function, under a standardized OMB process. [2]
The third ground echoes the 2017 bill’s compensating controls. The law also required GAO to report every two years on the waiver process, recommended practices and the waivers granted. [2] Those reports became the main public record of how the exceptions worked, as we discuss in the hindsight section. [5] [6]
The visibility: inventories, researchers and the database
The fourth part concerns information. A rule about devices is only as good as the government’s knowledge of which devices it has and what is wrong with them. The bill addressed that in three ways. [1]
Inventories
Each executive agency was to establish and maintain an inventory of internet-connected devices procured under the Act, within 180 days of enactment. OMB, with DHS, was to issue guidelines for those inventories based on DHS’s Continuous Diagnostics and Mitigation program. [1]
The inventory covered devices procured under the Act, not every connected device already in use. In our reading, that scope would have left legacy devices outside the list, and those are the devices least likely to meet the baseline.
Inventory as the foundation
Every other part of the bill depended on the inventory. Notifications are useless if the agency cannot tell which devices are affected. End-of-support dates are useless if the agency does not know where the device sits. Exceptions cannot be reviewed if no one lists them. [1]
The bill tied the inventory to DHS’s Continuous Diagnostics and Mitigation program, which agencies already used for other assets. [1] In our reading, that was a sensible choice of plumbing. It would put devices into the same discovery and reporting stream as other assets, instead of a separate list maintained by hand.
Researchers as a source
The bill directed DHS’s National Protection and Programs Directorate to issue, within 60 days of enactment, guidelines on coordinated disclosure for contractors providing devices to the government. They were to be based in part on ISO 29147, covering how contractors receive and resolve vulnerability reports. [1]
The guidelines were to require that research be done on the same class, model or type of device sold to the government, not on the actual device in government use. [1] The bill paired that with liability protection under the Computer Fraud and Abuse Act and the Digital Millennium Copyright Act for good-faith research that followed the guidelines. [1]
In our reading, these provisions treated outside researchers as part of the government’s detection architecture. Researchers find flaws in commercial devices. A legal shield for them, tied to devices the government buys, would increase the flow of findings to vendors and, through the notification clause, to agencies. [1]
Public lists as a market signal
The bill asked OMB to keep two public databases: one of devices and manufacturers covered by the research protections, and one of devices whose manufacturers had formally notified the government that security support had ceased. OMB was to update both at least every 30 days. [1]
In our reading, the second list would have reached far beyond federal buyers. A county, a hospital or a bank could check it before buying the same device. A manufacturer whose product appeared on it would face pressure from every buyer as well as the government. The bill used federal purchasing to publish information the whole market could use.
The national database
The bill’s last section directed NIST to establish, maintain and use best practices in identifying and tracking vulnerabilities for the National Vulnerability Database. [1] Because the baseline certification pointed to that database, its quality became part of the procurement rule. A gap in the database would become a gap in the certification. [1]
What the 2020 law did with visibility
The enacted law did not include the 2017 bill’s research liability protections or its public databases. It focused on agency standards, disclosure guidelines and contractor compliance. [1] [2] Separately, CISA’s Binding Operational Directive 20-01 required each agency to publish a vulnerability disclosure policy and maintain procedures for handling reports. [3]
The directive described a disclosure policy as committing the agency to authorize good-faith security research and respond to reports. [3] In our reading, it delivered part of what the 2017 bill’s research provisions aimed at, for agencies’ own systems, by directive rather than by amending criminal and copyright law.
Inventories came back through OMB guidance. GAO’s reports describe OMB requirements, issued in 2022 and 2023 and updated in January 2025, for agencies to establish IoT inventories and process waivers. [5] [6] The 2017 bill’s inventory clause had anticipated the need. [1]
How the approach reached different organizations
The bill was about federal purchasing, but its effects would reach well past federal buyers. Each kind of organization would meet it differently.
Federal agencies
For agencies, the bill would have added contract clauses, waiver petitions, inventories and research guidelines to the acquisition process. [1] Facilities, security, medical and laboratory equipment would all fall within the device definition if connected and processing data. In our reading, the hardest part would have been the device definition’s reach into equipment that program offices, not IT offices, may buy.
Device vendors
Vendors selling to the government would have had to certify the four baseline promises, notify of new flaws and support updates for the contract’s term. [1] A vendor building one product line for all markets could apply the same design to commercial units. The federal rule could shape devices sold to everyone.
Resellers and integrators
Agencies can buy devices through resellers and systems integrators rather than from manufacturers. The bill’s clauses applied to the contractor providing the device. [1] In our reading, a reseller asked to certify the four promises would need the same assurances from the manufacturer, flowed down by contract, because it could not verify firmware itself.
The 2020 law addressed the chain more directly in its disclosure provisions, which covered contractors and subcontractors at any tier. [2] That language recognized that the party signing the contract is not always the party that can fix the flaw.
Agency chief information officers
Under the 2017 bill, the CIO’s role was indirect. The clauses lived in contracts, the waivers went to OMB, and the inventory was a duty of the agency head, with no role named for the CIO. [1] Under the 2020 law, the CIO became the decision point: the official whose contract review could trigger the procurement prohibition and whose determination a waiver required. [2]
In our reading, that shift changed the CIO’s exposure. A CIO who clears a contract for a non-compliant device, or makes the determination behind a waiver, owns that decision on the record. That is a stronger accountability design than a vendor certification held in a contract file. It also depends on the CIO seeing every device purchase, including those made by facilities and program offices. [2] [5]
Contracting officers
The bill would have put new clauses in front of contracting officers, who are not security engineers. They would need to know which purchases involved internet-connected devices, which clauses applied and when a waiver petition was needed. [1] In our reading, the clause approach depended on acquisition staff spotting devices inside larger purchases, such as a building renovation that includes networked controls.
State and local governments
The bill did not apply to state or local agencies. [1] But they buy similar devices from the same vendors. In our reading, a state or county could adopt the bill’s four promises as its own contract clauses without new law, and gain the same baseline.
Hospitals
A health system’s connected equipment can range from clinical devices to building systems. The bill did not reach hospitals. [1] Its lifecycle clauses, especially the end-of-support notice, describe what a hospital would want from a device vendor. Clinical devices raise further questions of their own, which the bill did not address.
Banks
Banks can rely on connected devices too, in branches, building controls and physical security systems. The bill did not apply to them. [1] A bank’s third-party risk program could still borrow its clauses, especially the ban on hard-coded credentials and the requirement for authenticated updates.
Federal contractors outside IT
The device definition did not depend on what a contractor called itself. A construction firm installing networked building controls, a laboratory supplier delivering connected instruments or a fleet vendor providing telematics would each be providing internet-connected devices under the bill’s definition, if the devices met it. [1]
In our reading, those contractors would have faced a steep adjustment. They sell equipment, not security, and the clauses would have asked them to certify facts about firmware and credentials that their own suppliers control. Their supply chains run to component makers who may never have been asked about hard-coded credentials or signed firmware. The bill would have reached them through their federal contracts, and through them, their suppliers. [1]
Security researchers
The research protections would have applied only to devices of a class, model or type sold to the government, and only for research following the guidelines. [1] That scope tied researchers’ legal safety to government purchasing, a narrower protection than a general research exemption.
Objections, in order
Our reading has reasonable critics. We take the strongest objections in turn.
1. A bill that did not pass tells us little
S. 1691 was not the bill that became law. A later bill did. The 2020 law kept the bill’s purpose and several of its ideas, including coordinated vulnerability disclosure and waivers for devices that cannot comply. [1] [2] The 2017 text shows the design choice the 2020 law changed.
2. Vendor certifications are the normal way to buy
Certifications are a standard tool of federal contracting. The bill used one sensibly. Our point is that a certification places the test with the seller at a single moment, while device risk changes over the device’s life. [1] The enacted law placed the test with agency officials reviewing contracts against NIST standards. [2]
3. The baseline was too weak to matter
Four promises do not make a device secure. But each targets a flaw that design can avoid. [1] A baseline is meant to screen out avoidable flaws, not to certify strong security.
4. The exceptions would swallow the rule
The limited-functionality waiver, the third-party standards route and agency evaluation processes each offered a way around the baseline. [1] That risk was real. The bill answered it with OMB approval, NIST alignment and the option of deadlines for removing non-compliant devices. GAO’s later findings on how agencies reported waivers under the 2020 law suggest that oversight of exceptions is harder than writing them. [5]
5. Network controls make device security unnecessary
Segmentation and gateways can protect weak devices. [1] The bill treated them as compensating controls for exceptions, not as a substitute for the baseline. In our reading, that ordering is right: network controls fail too, and a device with no hard-coded credentials is safer behind any network.
6. Certifications would produce paperwork, not security
A certification can become a form a vendor signs without checking. That is a fair concern about any self-attestation. The bill offered some safeguards: third-party certification for the equivalent-standards route, NIST accreditation of certifiers, and the government’s contractual remedies. [1] The enacted law’s choice to place the decision with agency CIOs reflects the same concern from the other side. [2]
7. Federal purchasing is too small a lever to change devices
Federal agencies are one buyer in a large market. A vendor might choose not to sell to them rather than redesign a product. The bill did not depend on changing every device. It aimed to change the devices on federal networks, and its public end-of-support database could inform other buyers. [1] The bill’s five-year report to Congress on the guidelines’ effectiveness, with recommended legislative updates, is where an answer to that question could have surfaced. [1]
What an agency architect could check
The bill’s four parts translate into questions any organization can ask about its connected devices, whether or not a law requires it.
- Does every connected device in the inventory have a named owner and a recorded end-of-support date?
- Do purchase contracts ban fixed or hard-coded credentials for administration, updates and communication?
- Can each device accept only signed updates, and who holds the process for applying them?
- Which devices cannot meet the baseline, and what segmentation or gateway isolates each one?
- Who approves an exception, in writing, and when is it reviewed?
The questions can be asked of any connected device estate, federal or not. A hospital can ask them of its clinical engineering inventory. A bank can ask them of its branch and facilities equipment. A county can ask them before its next purchase of cameras or sensors. [1]
The first and fifth questions are the ones GAO’s reports on the 2020 law suggest agencies struggled with. The middle three come straight from the 2017 bill’s baseline and exceptions. [1] [5] [6] An organization that can answer all five has done what the bill asked of federal agencies, by its own choice.
With hindsight
The law that was enacted came from a later bill. H.R. 1668 became the IoT Cybersecurity Improvement Act of 2020, approved December 4, 2020. [2] The enacted law took a different route. Instead of contract clauses certified by vendors, it directed NIST to publish standards and guidelines for agencies’ use and management of IoT devices, including minimum security requirements. [2]
The law tied those standards to agency practice. OMB was to review agency security policies for consistency with them, and the Federal Acquisition Regulation was to be revised as necessary. [2] NIST was to publish guidelines for reporting, coordinating and receiving vulnerability information, aligned to ISO standards 29147 and 30111, including for contractors and subcontractors at any tier. [2]
The central enforcement provision moved the test to the agency. Two years after enactment, an agency head became prohibited from procuring, renewing or using an IoT device that prevented compliance with the NIST standards or disclosure guidelines. The finding was the CIO’s, made in a required contract review. [2] The agency head could waive the prohibition where the CIO determined it was needed for national security or research, or that the device was secured by alternative and effective methods, under a standardized OMB process. [2]
NIST published SP 800-213, IoT Device Cybersecurity Guidance for the Federal Government, in November 2021. It helps organizations identify the device cybersecurity requirements they will expect from a device and its manufacturer, in the context of system risk management. [4] The vulnerability disclosure strand had its own path. In September 2020, CISA issued Binding Operational Directive 20-01, requiring each agency to publish a vulnerability disclosure policy and maintain supporting handling procedures. [3]
Implementation has been slow, by GAO’s account. In December 2024, GAO reported that many of the 23 civilian agencies covered by the law had not fully addressed OMB’s requirements on inventories and waivers. [5] Six agencies had reported IoT waivers, but officials at five said they should not have, and OMB had not verified the waiver data. [5]
GAO’s final report under the law, published September 30, 2026, found that as of that month, 15 of 22 civilian CFO Act agencies had established an inventory of networked devices. Eleven were maintaining them, ten included all required information, and seven had fully addressed all three OMB requirements. [6] No agency had reported an IoT cybersecurity waiver. GAO found OMB had yet to issue updated guidance covering fiscal year 2026. [6]
What held up from the 2017 bill? Its four baseline promises still describe good device design, and NIST’s later guidance treats device capabilities as requirements an organization should expect. [1] [4] Its inventory provision anticipated the requirement GAO found agencies still meeting unevenly in 2026. [1] [6]
What did not hold was the bet on vendor certification as the main control. The enacted law put the decision with agency CIOs reviewing contracts against standards. [2] GAO’s findings suggest that this route depends on something the 2017 bill also needed and also struggled to secure: agencies knowing what devices they have. [1] [6]
In our reading, the inventory is the part of the architecture both versions relied on and neither could supply by statute alone. GAO’s 2026 report cited a July 2026 disruption in the water sector, where threat actors modified passwords to disconnect networked controllers. [6] Devices that are not inventoried cannot be checked against any standard, whoever writes it.
Spotted an error? Tell us and we will check it and correct the article where needed.