FDA’s January 2016 device cyber draft let a notice telling hospitals how to cut the risk count as remediation
The postmarket draft announced January 15, 2016 defined remediation to include a customer notice with a risk-reduction strategy, such as taking equipment off the network. That made a hospital's ability to act on notices part of patient safety.
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 postmarket draft announced January 15, 2016 defined remediation to include a customer notice with a risk-reduction strategy, such as taking equipment off the network. That made a hospital's ability to act on notices part of patient safety.
- The week The release explained why a postmarket document was needed.
- The split: controlled and uncontrolled risk The draft gave an example of why the definition matters.
- The routine patch: an enhancement, not a repair The release spoke directly to whether security changes needed FDA action.
- The notice as remediation: what users were asked to do If a notice can be the remediation, its quality is a safety matter.
- The reporting bargain: 30 days and an ISAO The draft measured its 30 days from when the manufacturer learns of the vulnerability. [2] That makes intake a control.
- How the draft reached different organizations A system with a clinical engineering department and a security operations team could build the capabilities the notice model assumed.
- Objections, in order The draft’s recommendations ran to manufacturers.
- Checks a hospital could run Keep an inventory that ties each networked device to its model, software version, network location and clinical owner.
- With hindsight The final kept the controlled and uncontrolled split but renamed its anchor.
The week
On January 15, 2016, the Food and Drug Administration announced a draft guidance on cybersecurity for medical devices already on the market. [1] We read the release in an Internet Archive copy of the agency’s newsroom page, captured on January 17, 2016. [1]
The release said the document outlined steps manufacturers should take to continually address cybersecurity risks. [1] It said the guidance details the agency’s recommendations for monitoring, identifying and addressing cybersecurity vulnerabilities in medical devices once they have entered the market. [1]
The release explained why a postmarket document was needed. It said manufacturers can build controls into a product’s design, but must also consider improvements during maintenance, because risks may arise throughout a device’s entire lifecycle. [1]
It said the guidance recommends that manufacturers implement a structured and systematic comprehensive cybersecurity risk management program and respond in a timely fashion to identified vulnerabilities. [1] It listed seven critical components of such a program. [1]
The first was applying the 2014 NIST voluntary Framework for Improving Critical Infrastructure Cybersecurity. [1] The release named its core principles as “Identify, Protect, Detect, Respond and Recover.” [1] The others were monitoring cybersecurity information sources and understanding, assessing and detecting the presence and impact of a vulnerability. [1] They also included establishing and communicating processes for vulnerability intake and handling. [1]
The remaining components were clearly defining essential clinical performance, adopting a coordinated vulnerability disclosure policy and practice, and deploying mitigations early and prior to exploitation. [1] It said the guidance addresses the need to plan for and assess vulnerabilities, consistent with the agency’s Quality System Regulation. [1]
It also addressed information sharing. The release said the guidance addresses participation in an Information Sharing Analysis Organization (ISAO), which it described as a collaborative group in which public and private-sector members share cybersecurity information. [1]
The release then described the reporting position. It said that for the majority of cases, manufacturers’ actions on vulnerabilities and exploits are considered “cybersecurity routine updates or patches.” [1] For those, it said, the agency does not require advance notification, additional premarket review or reporting under its regulations. [1]
It described a narrower class: a small subset that may compromise a device’s essential clinical performance. [1] Where such a flaw presents a reasonable probability of serious adverse health consequences or death, the release said manufacturers would be required to notify the agency. [1]
The release then set out conditions under which the agency did not intend to enforce urgent reporting of such a vulnerability. [1] There must be no serious adverse events or deaths associated with it. [1] Within 30 days of learning of it, the manufacturer must notify users and implement changes that reduce the risk to an acceptable level. [1]
The third condition was information sharing. The release said the manufacturer must be a participating member of an ISAO and report the vulnerability, its assessment and remediation to the ISAO. [1]
The release said comments would be open for 90 days. [1] It said the agency would discuss the guidance at a public workshop, “Moving Forward: Collaborative Approaches to Medical Device Cybersecurity,” on January 20-21 at its headquarters in Silver Spring, Maryland. [1]
It placed the guidance in a longer effort. The release said the agency had been working on cybersecurity information sharing and risk-based standards since 2013, when the White House issued Executive Order 13636 and Presidential Policy Directive 21. [1] It said that in October 2014 the agency finalized its guidance on premarket management of cybersecurity during device design. [1]
The release listed other steps. It named formal partnerships with the Department of Homeland Security’s Industrial Control Systems Cyber Emergency Response Team and the National Health Information Sharing and Analysis Center, and a 2014 public workshop. [1] It also mentioned product-specific safety communications on medical device cybersecurity vulnerabilities. [1]
The January 2016 draft can be read as the FDA asking device makers to patch. [1] That reading misses where the draft placed part of the fix. Its definition of remediation included a notice to the customer base that discloses a vulnerability and gives a strategy to reduce the risk to an acceptable level. [2]
Our argument is that the draft made the hospital’s network and staff part of a device’s remediation. A manufacturer could meet the draft’s 30-day path with compensating controls carried out by users, such as taking a device off the hospital network. [2] That turned a hospital’s ability to receive and act on a device notice into a safety function.
The draft was written for manufacturers, and its duties ran to them. [2] Its effect on hospitals came through definitions and examples rather than direct requirements. That is why it is easy to read past. A hospital reading for its own obligations would find few. A hospital reading for what the manufacturer would expect of it would find a good deal.
That gap between duty and dependence is the subject of this piece. The sections below follow a vulnerability from its rating to its report, and mark at each step what the hospital was assumed to do.
We take the draft in four parts, each answering a different question about a vulnerability:
- The split: how the draft sorted vulnerabilities into controlled and uncontrolled risk.
- The routine patch: what the draft said about updates for controlled risk.
- The notice as remediation: what the draft asked users to do, and what that asked of hospitals.
- The reporting bargain: the conditions for relief from reporting, including the ISAO.
The parts do not overlap. The first is the classification, the second covers the low-risk branch, the third covers who carries out the fix on the high-risk branch, and the fourth covers what the FDA is told. After them we look at how the draft reached different organizations, take the objections in order, and end with checks a hospital could run.
One note on dates. The release is dated January 15, 2016. [1] The draft’s cover says “Document issued on: January 22, 2016,” the day the Federal Register notice of availability was published. [2] [3] We read the draft in an Internet Archive copy captured on January 19, 2016. [2]
The split: controlled and uncontrolled risk
Essential clinical performance
The draft built its classification on a term it said it developed for the purpose: essential clinical performance. [2] It defined it as performance necessary to achieve freedom from unacceptable clinical risk, as defined by the manufacturer. [2] It said compromise of that performance can produce a hazardous situation that results in harm or may require intervention to prevent harm. [2]
The draft gave an example of why the definition matters. It said a vulnerability affecting the essential clinical performance of a thermometer may be quite different from one affecting an insulin infusion pump. [2] Acceptable mitigations, it said, will vary accordingly. [2]
The phrase “as defined by the manufacturer” carries weight. [2] The hospital does not set the threshold. The manufacturer’s own risk file decides what counts as essential, and so what counts as a serious vulnerability.
Criteria set before the flaw
The draft asked manufacturers to decide their thresholds before a vulnerability arrives. It said they should define, as part of risk management, the essential clinical performance of their device, the resulting severity outcomes if compromised, and the risk acceptance criteria. [2] It said defining these allows manufacturers to triage vulnerabilities for remediation. [2]
That sequence protects against a known failure in incident handling: setting the bar after seeing the cost of clearing it. A manufacturer that writes its acceptance criteria in advance can be held to them. A hospital that asks to see those criteria can judge whether a notice is consistent with them.
Two measures
The draft asked manufacturers to assess risk to essential clinical performance through two factors: the exploitability of the vulnerability and the severity of the health impact if it were exploited. [2] It said the analysis should also consider compensating controls and risk mitigations. [2]
For exploitability, the draft noted that estimating the probability of an exploit is difficult. [2] It suggested a scoring system such as the Common Vulnerability Scoring System, Version 3.0, and listed its factors, from attack vector to report confidence. [2]
For severity, it pointed to the qualitative levels described in ANSI/AAMI/ISO 14971: negligible, minor, serious, critical and catastrophic. [2] The draft described catastrophic as resulting in patient death. [2]
Two scores, two owners
The two measures have different owners. Exploitability can be scored from outside: a public advisory may publish a CVSS score that anyone can read. Severity of health impact depends on the device’s clinical use, which the manufacturer defines. [2]
A hospital security team that sees only the CVSS score sees half of the draft’s assessment. A high exploitability score on a device whose essential clinical performance is not affected can rate as controlled. A modest score on an infusion pump can rate as uncontrolled. The hospital needs the manufacturer’s severity view to rank its own work.
A binary answer
The draft then asked for a decision. It said manufacturers should make a binary determination that a vulnerability is either controlled or uncontrolled, using an established process tailored to the product. [2] It acknowledged that in some situations the determination may not be distinct. [2]
Controlled risk, in the draft, is present when residual risk to essential clinical performance is sufficiently low. [2] Uncontrolled risk is present when residual risk is unacceptable due to insufficient compensating controls and risk mitigations. [2]
That definition of uncontrolled risk refers to compensating controls. In our reading, it means the classification can change without any change to the device. A compensating control put in place by users can move a vulnerability from uncontrolled to controlled. The draft’s own examples show it happening. [2]
Device features first
The draft did not make users the first line. Its appendix said the FDA recommends that manufacturers implement device-based features as a primary mechanism to mitigate the impact of a vulnerability. [2] It said manufacturers should also assess and prescribe compensating controls to users, so the risk is further mitigated by a defense-in-depth strategy. [2]
That is the strongest point against our argument, and we take it seriously. The draft’s preference ran to the device. Its definitions and examples still allowed a user-applied control to carry the whole remediation when no device fix existed. [2] The preference described the ideal. The examples described what would earn reporting relief.
What the classification decides
The classification sends a vulnerability down one of two paths. Controlled risk leads to routine updates and, for some devices, an annual report. [2] Uncontrolled risk leads to remediation, user communication and a reporting decision. [2]
For a hospital, the classification is invisible unless the manufacturer shares it. A security bulletin that describes a flaw without saying how the manufacturer rated it leaves the hospital to guess which path the device is on. The draft’s recommended content for PMA periodic reports included whether the risk was controlled or uncontrolled. [2] That report goes to the FDA, not to customers.
The routine patch: an enhancement, not a repair
What the draft defined
The draft defined “cybersecurity routine updates and patches” as changes to increase device security or remediate vulnerabilities associated with controlled risk. [2] They are not made to reduce a risk to health or correct a violation of the FD&C Act. [2]
It said such updates are generally considered a type of device enhancement and not a repair. [2] It said they include regularly scheduled security updates, and updates for controlled-risk vulnerabilities made earlier than their scheduled cycle, even if distributed to multiple units. [2] Labeling changes that strengthen cybersecurity through user education can also count. [2]
The draft said these changes are generally not required to be reported under 21 CFR part 806, the FDA’s rule on reports of corrections and removals. [2] It pointed to the FDA’s guidance on distinguishing recalls from enhancements. [2]
Why the distinction mattered
The release spoke directly to whether security changes needed FDA action. It said the FDA does not require advance notification, additional premarket review or reporting for routine updates. [1]
The draft was more qualified. It said that for routine updates the FDA will not, in the typical case, need to conduct premarket review to clear or approve the software changes. [2] It also said manufacturers should evaluate device changes to assess whether a premarket submission is needed. [2] The relief was real but conditional on the manufacturer’s own change assessment.
What the label changes
The enhancement label has a practical effect on records. Part 806 covers reports of corrections and removals to the FDA. [2] A routine cybersecurity update does not need that report, under the draft’s reading. [2] For a hospital, that means a routine security update may arrive with no FDA report behind it. The hospital’s own change records become the main evidence that the update was applied.
That places a premium on the hospital’s configuration records. If an auditor or a patient safety review later asks which software version was running on a device on a given date, the answer will come from the hospital’s records, not from a public database.
What the manufacturer still owed
The draft kept quality system duties in place for routine updates. It said manufacturers should conduct software validation under 21 CFR 820.30(g) so that a remediation does not create new exposure. [2] It said they should document methods and controls under 21 CFR part 820. [2]
For Class III devices with PMA periodic reporting, the draft said cybersecurity changes made as routine updates should go into the annual report. [2] Its example was malware on the PC component of a Class III device that collected browsing data without affecting clinical performance. [2] The draft said the update need not be reported under 806.10, but should appear in the annual report. [2]
Labeling as a patch
The definition of routine updates also covered changes to product labeling, including instructions for use, to strengthen cybersecurity through user education and best practices. [2] In our reading, that means a revised manual can be a cybersecurity change. A hospital that files the old manual and never sees the new one can miss a control the manufacturer considers part of the device.
Document control for device manuals belongs with the same team that tracks software versions. A notice that changes the instructions for use is a configuration change for the hospital, even if no code moves.
The hospital’s view of routine patches
The draft removed one barrier and left others. A hospital still depends on the manufacturer to validate and release a patch. The draft did not set a time frame for routine updates. A device with a controlled-risk flaw can stay unpatched for as long as the manufacturer’s schedule allows, with the hospital carrying the residual risk.
That residual risk is not zero. Controlled means acceptable to the manufacturer’s criteria for essential clinical performance. [2] It says less about the device’s role as a foothold on the hospital network. The draft’s definition of an exploit did include using a device as a vector to compromise the performance of a connected device or system. [2]
The notice as remediation: what users were asked to do
The definition
The draft defined remediation as any action taken to reduce the risk to a device’s essential clinical performance to an acceptable level. [2] It said remediation may include complete solutions that remove a vulnerability, or compensating controls that adequately mitigate the risk. [2]
It then gave an example of remediation: a notification to the customer base and user community. [2] The notification discloses the vulnerability and its potential impact, and provides a strategy to reduce the risk to an acceptable level. [2] It added that if the notification does not provide such a strategy, the remediation is considered incomplete. [2]
What a usable notice contains
If a notice can be the remediation, its quality is a safety matter. The draft asked that users receive relevant information on work-arounds, temporary fixes and residual cybersecurity risks. [2] It did not set a format.
From a hospital’s side, a usable notice names the affected models and software versions, says what the vulnerability allows, states the recommended control and how to verify it is in place, and says what risk remains. It also says whether a fix is planned. A notice missing any of these leaves hospital staff to fill the gap by guesswork, at the moment the manufacturer is relying on them.
The draft’s definition contains its own test. A notification without a strategy to reduce the risk to an acceptable level is incomplete remediation. [2] A hospital can apply that test to every notice it receives and ask for the missing strategy in writing.
Updates designed in
The draft also looked ahead to design. In its appendix, it said manufacturers should design devices so that risks inherent in remediation are mitigated. [2] It listed ensuring the remediation is adequate and validated, and that device designs incorporate mechanisms for secure and timely updates. [2]
A device that can take a secure, timely update needs less from the hospital when a flaw appears. A device that cannot puts the weight on compensating controls. The hospital’s purchasing decision therefore shapes how much of a future remediation its own staff will carry.
Compensating controls
The draft defined a compensating control as a safeguard external to the device, employed by a user in place of, or in the absence of, sufficient controls designed in by the manufacturer. [2] It said the definition was adapted from NIST Special Publication 800-53A. [2]
Its example was a device that can operate safely without the hospital network. The manufacturer instructs users to configure the network to remove unauthorized or unintended access to the device. [2] The draft called that a compensating control. [2]
The draft’s own example
One of the draft’s uncontrolled-risk examples follows that path. A Class II device carries a vulnerability known to the security community but not to the manufacturer at development. [2] No failures or injuries are reported, but the vulnerability introduces a failure mode affecting essential clinical performance. [2]
The manufacturer has no software update. It notifies the customer base and instructs them to disconnect the device from the hospital network. [2] The draft said the company’s risk assessment concludes the risk is then controlled. [2] If that happened within 30 days and the company was an ISAO member, the FDA did not intend to enforce the 806 reporting requirement. [2]
In that example, the hospital performs the whole remediation. The manufacturer’s part is the notice.
What that asks of a hospital
For the notice to work as remediation, a hospital needs four things. It needs to receive the notice at an address someone reads. It needs to know which of its devices are affected, by model and software version. It needs network staff able to isolate those devices. And it needs clinical staff who can run the device without the network, if the workflow depends on it.
Each is an operational capability with an owner. Device inventories sit with clinical engineering. Network segmentation sits with IT. Workflow changes sit with nursing and the clinical departments. The draft’s model assumed those owners coordinate within the manufacturer’s 30-day window. [2]
Finding every affected device
The draft asked manufacturers to assess impact across their products. Its appendix described analysis across a portfolio, sometimes called variant analysis, and within a device’s components. [2] It said a vulnerability found in one device may affect others, including those in development or not yet cleared, approved or marketed. [2]
Hospitals face the same problem from the other side. One component flaw can appear in devices from several manufacturers. Each sends its own notice on its own schedule. The hospital is the only party that sees all of them land on its own network.
The cost of disconnection
Taking a device off the network is a safety measure with clinical costs. Results may have to be transcribed by hand. Alarms may not reach central stations. Orders may have to be keyed at the bedside. The draft recognized the decision belongs with users who have information. It said users should get information on work-arounds, temporary fixes and residual risks so they can make informed decisions about device use. [2]
In our reading, a hospital that cannot carry out a compensating control is left with an uncontrolled risk that the manufacturer’s records may show as controlled. The classification assumed the instruction was followed.
Detection
The draft also noted a limit on the device side. It said medical devices may not be capable of detecting threat activity and may rely on network monitoring. [2] It encouraged manufacturers to consider design features that support detection and forensically sound evidence capture. [2]
Until such features exist, detection is the hospital’s job. The draft’s model of shared responsibility named health care facilities among the stakeholders. [2] The monitoring that sees an attack on a device is, under that model, a hospital capability.
The reporting bargain: 30 days and an ISAO
The conditions in the draft
For uncontrolled risk, the draft said manufacturers should report the vulnerabilities under 21 CFR part 806 unless reported under parts 803 or 1004. [2] It then said the FDA did not intend to enforce part 806 reporting if three circumstances were all met. [2]
First, there are no known serious adverse events or deaths associated with the vulnerability. [2] Second, within 30 days of learning of it, the manufacturer identifies and implements device changes or compensating controls to bring residual risk to an acceptable level, and notifies users. [2] Third, the manufacturer is a participating member of an ISAO, such as NH-ISAC. [2]
When the clock starts
The draft measured its 30 days from when the manufacturer learns of the vulnerability. [2] That makes intake a control. A manufacturer that has no clear route for researchers or customers to report a flaw may learn of it late, and its clock starts late.
The draft’s appendix asked for a clear, consistent and reproducible process for intake and handling of vulnerability information. [2] It pointed to ISO/IEC 29147:2014 on vulnerability disclosure, as well as ISO/IEC 30111:2013 on handling, as resources. [2] For a hospital that finds a flaw in a device, the manufacturer’s intake channel is where the clock begins. Reporting through that channel, and keeping the date, gives the hospital its own record of when the manufacturer knew.
The difference between the release and the draft
The release described the third condition as membership plus reporting the vulnerability, its assessment and remediation to the ISAO. [1] The draft’s text of the condition named membership. [2] The draft’s annual report list asked for the date and name of the ISAO to which a vulnerability was reported, if any. [2] A reader relying on the release alone would have read the bar as slightly higher than the draft’s wording.
Why tie relief to an ISAO
The draft explained its view. It said the agency considers voluntary ISAO participation a critical component of a manufacturer’s proactive approach to postmarket threats. [2] It said the CDRH had entered into a memorandum of understanding with NH-ISAC to encourage sharing of information about threats and vulnerabilities. [2]
The draft cited Executive Order 13691, released on February 13, 2015, which encourages ISAOs as focal points for sharing. [2] It described protections for shared information under the Critical Infrastructure Information Act of 2002, where its requirements are met. [2]
What the hospital gets from the bargain
The reporting bargain is between the manufacturer and the FDA. A hospital’s benefit is indirect. If the manufacturer shares with an ISAO that the hospital also joins, the hospital may learn of a vulnerability through that channel as well as through the manufacturer’s notice.
The draft did not require the manufacturer to tell the hospital how it classified the risk, or whether it filed an 806 report. A footnote said manufacturers and user facilities may have additional reporting requirements from sources other than the FDA. [2] For a hospital, those may include its own breach and incident duties.
What the FDA would see
For PMA devices with periodic reporting, the draft listed recommended content for the annual report. [2] The list included a description of the vulnerability and how the firm learned of it, and a summary of the risk assessment, including whether the risk was controlled or uncontrolled. [2]
It also included a description of the changes, the rationale, related submissions, any Unique Device Identification, a link to an ICS-CERT advisory if applicable, and the ISAO date and name. [2] In our reading, the annual report gave the FDA a record of the manufacturer’s decisions after the fact. The ISAO channel and the customer notice were the paths that reached others while the risk was live.
The serious adverse event limit
The relief ended where harm began. The draft’s last example involved a hospital report that a patient was harmed after a device malfunctioned because of an exploited vulnerability, which the draft said may have contributed to a patient death. [2] The manufacturer notified the customer base and user community, developed a validated emergency patch within 30 days of learning of the vulnerability, and was an ISAO member. [2] Because there was a serious adverse event or death, the draft said it files a report under 806.10 and complies with part 803. [2]
How the draft reached different organizations
A large health system
A system with a clinical engineering department and a security operations team could build the capabilities the notice model assumed. The gap could sit between them: device inventories kept in one tool and network inventories in another. The draft’s model rewarded a single view of each device, its software version and its network location.
An ambulatory network or physician group
Smaller sites own fewer devices but have less staff to act on a notice. A clinic with no network engineer cannot segment a device in days. For such sites, the compensating control may be physical disconnection, with the workflow cost that brings. Shared services or a managed provider can close part of that gap.
A device manufacturer
For manufacturers, the draft created a choice. Fix within 30 days and join an ISAO, or file 806 reports. [2] The choice favored investment in vulnerability intake, assessment and customer communication. The draft’s appendix listed intake processes and pointed to ISO/IEC 30111:2013 on vulnerability handling as a resource. [2]
A health IT or software vendor
The draft’s scope covered devices containing software, firmware or programmable logic, and software that is itself a medical device. [2] It said it did not apply to investigational devices. [2] It listed Health IT developers among the stakeholders, alongside IT vendors that provide products not regulated by the FDA. [2] Those vendors’ products sit on the same networks without the same FDA framework.
A security operations team
For a hospital’s security operations team, the draft added a class of assets it could see but not patch. Network monitoring can flag a device’s unusual traffic. The response options are limited to the network: isolate, block or watch. The playbook for a medical device alert therefore differs from the playbook for a server, and clinical engineering has to be on the call before anything is isolated.
A clinical engineering contractor
Some hospitals outsource device maintenance to independent service organizations. The draft’s notice model assumed a path from manufacturer to user. Where a contractor maintains the device, the notice may need to reach the contractor as well, and the service contract should say who acts on it. Without that, a notice can arrive at a department that no longer touches the device.
A public hospital or government health program
Public hospitals buy and run the same devices and face the same notice model. They also face procurement rules that can slow a network change or a replacement purchase. The 30-day clock belonged to the manufacturer. A public hospital’s own timeline for acting on a notice can be set by procurement as much as by clinical risk.
Objections, in order
1. Manufacturers, not hospitals, are responsible for device safety
The draft’s recommendations ran to manufacturers. It did not change who answers to the FDA. It did define remediation to include steps users take. [2] The objection describes accountability to the regulator. It does not describe who carries out a compensating control.
2. The FDA said routine updates cover the majority of cases
The release said so. [1] That may hold for clinical risk as the manufacturer defines it. The hospital also carries enterprise risk: a controlled-risk device can still be a path into other systems. The draft’s definition of an exploit included that path. [2]
3. A 30-day clock is fast
The 30-day condition bound the manufacturer’s reporting relief, not the hospital. A hospital that cannot act on a notice within that window does not lose anything with the FDA. Its patients carry the residual risk.
4. ISAO membership is a formality
The draft tied relief to membership. [2] Membership alone does not guarantee sharing. The release described reporting to the ISAO. [1] The final guidance tightened this, as the next section shows. [4]
5. Hospitals can refuse to buy insecure devices
Purchasing is a real lever. It works on new devices. It does little for the installed base, which a hospital may run for years under service contracts. The draft addressed marketed devices, the ones already in service. [2] The notice model mattered because replacement is slow.
6. Disconnecting is always the safe choice
Removing network access can remove the attack path. It can also remove alarms, data flows and remote monitoring that patients rely on. The draft asked manufacturers to tell users about residual risks so they can decide on device use. [2] The decision weighs one clinical risk against another.
7. Guidance is not binding
The draft said guidance documents do not establish legally enforceable responsibilities, unless specific regulatory or statutory requirements are cited. [2] The requirements it cited, part 806 and part 820, are binding. The guidance described how the FDA would exercise discretion over them.
Checks a hospital could run
Know the devices
Keep an inventory that ties each networked device to its model, software version, network location and clinical owner. Without it, a notice cannot be matched to devices within a manufacturer’s 30-day window. [2]
Route the notices
Name a mailbox and an owner for device security notices from each manufacturer. Test that a notice reaches clinical engineering and IT security within one business day.
Rehearse a compensating control
Pick one device class and rehearse taking it off the network. Record what clinical workflow changes, who approves it and how long it takes.
Ask how the risk was rated
When a notice arrives, ask the manufacturer whether it rated the risk controlled or uncontrolled, and on what basis. The draft’s annual report list shows manufacturers record that rating. [2]
Watch the device on the network
Where a device cannot detect attacks, monitor its traffic. The draft said devices may rely on network monitoring. [2]
Join the sharing channel
Consider membership in a health sector ISAO, so that vulnerability information arrives through more than one path.
Report flaws through the front door
When hospital staff find a device vulnerability, report it through the manufacturer’s published intake channel and keep the date. The draft measured the manufacturer’s 30 days from when it learns of a vulnerability. [2]
Track the clock
When a notice arrives, record the date and the manufacturer’s stated timeline. Compare later notices against it. A record of each manufacturer’s timeliness is useful at contract renewal.
Give the board a device view
Report to the board on networked devices with open notices, the controls applied and the residual risk accepted. Treat that as a patient safety report as well as a security report.
Plan for no fix
For devices whose makers cannot issue a fix, keep a standing compensating control and an owner for it. Review it when the network changes, since a later network change can undo an isolation that was set up once.
Write it into contracts
Ask in purchasing contracts for vulnerability notices that include a remediation strategy, a timeline and the residual risk. The draft said users should receive information on residual risks to make informed decisions. [2]
With hindsight
The FDA finalized the guidance on December 28, 2016. [4] [5] The final document says the draft was issued on January 22, 2016. [4]
The final kept the controlled and uncontrolled split but renamed its anchor. Its headings refer to “safety and essential performance” in place of essential clinical performance. [4]
It changed the reporting bargain. For uncontrolled risk, the final said the FDA does not intend to enforce part 806 reporting when four circumstances are met. [4] The first remained no known serious adverse events or deaths. [4]
The second changed what the 30 days covered. In the final’s words, “as soon as possible but no later than 30 days after learning of the vulnerability,” the manufacturer communicates with customers, identifies interim compensating controls and develops a remediation plan. [4] It must document the timeline rationale for that plan. [4]
The final also listed minimum contents for that customer communication. It must describe the vulnerability and an impact assessment, describe compensating controls if any, and state that the manufacturer is working to fix it or provide a defense-in-depth strategy. [4]
A new third condition added a second clock. Within the final’s deadline, “as soon as possible but no later than 60 days after learning of the vulnerability,” the manufacturer fixes it, validates the change and distributes the deployable fix. [4] The final said a compensating control could in some circumstances produce a long-term solution, if the risk of patient harm is brought to an acceptable level. [4]
The fourth condition tightened the ISAO term. The manufacturer must actively participate as a member of an ISAO that shares vulnerabilities and threats affecting medical devices, and provide the ISAO with any customer communications. [4]
In our reading, the final moved part of the fix back to the manufacturer. A notice could still start the response, but reporting relief now required a deployable fix within 60 days, with a long-term compensating control as the exception the final described. [4]
Congress later wrote some of this into law. The FDA’s February 3, 2026 premarket guidance says section 3305 of the Food and Drug Omnibus Reform Act of 2022, enacted December 29, 2022, added section 524B to the FD&C Act. [6] It says that, effective March 29, 2023, sponsors of “cyber devices” must include cybersecurity information in premarket submissions. [6]
Section 524B(b)(1), as the guidance quotes it, requires a plan for postmarket vulnerabilities and exploits. [6] The quoted text is “a plan to monitor, identify, and address, as appropriate, in a reasonable time,” those issues, “including coordinated vulnerability disclosure and related procedures.” [6]
The guidance ties the statute back to the 2016 split. It says section 524B(b)(2)(B) requires updates and patches, as soon as possible out of cycle, for critical vulnerabilities that could cause uncontrolled risks. It says that, in general, this includes vulnerabilities that could cause uncontrolled risks as described in the Postmarket Cybersecurity Guidance. [6] It says updates for known unacceptable vulnerabilities on a regular cycle can include vulnerabilities that present controlled risk. [6]
The guidance page says the February 2026 document supersedes the version issued June 27, 2025. [7] The final postmarket guidance now opens with a note that the FDA amended 21 CFR part 820 into the Quality Management System Regulation, effective February 2, 2026. [4] As checked on November 16, 2026, the FDA’s page lists the postmarket guidance as final, dated December 2016. [8]
What held up is the split. The controlled and uncontrolled categories the January 2016 draft introduced now appear in the FDA’s reading of the statutory patch duties for cyber devices. [6] What changed is the weight on the notice. The 2016 final added a 60-day fix, and the 2022 law required plans and patches. [4] [6] Hospitals still carry out compensating controls. In our reading, a notice alone no longer earns full reporting relief. [4]
Spotted an error? Tell us and we will check it and correct the article where needed.