Medical device recalls for software and cybersecurity, 2016 to 2026
We counted FDA recall records and cybersecurity safety communications to see how often software and security flaws force device fixes, and how FDA's public warnings have changed.
The question and the two records
FDA keeps two public records that show when device software fails or is found exposed: recall records and a list of cybersecurity safety communications [1] [6]. We counted both from 2016 to see how often software causes a recall, how often a recall cites a cybersecurity weakness, and how FDA's public warnings have changed.
FDA uses "recall" for a manufacturer's correction or removal of a device that violates FDA law, and in most cases the company starts it on its own [8]. FDA classifies recalls by hazard: Class I where there is a reasonable chance of serious health problems or death, Class II where a product may cause a temporary or reversible problem, and Class III where harm is not likely [8].
Data and method
Recall counts come from the openFDA device recall endpoint, whose data were last updated November 12, 2026 when we queried it on November 16, 2026 [1]. openFDA draws it from FDA's Medical Device Recalls database and updates it weekly [2]. Each record is one recalled product, and products recalled together share a recall event number that FDA assigns for tracking [3]. We count distinct events, not products, and place each in the year of event_date_initiated, the date the firm first began notifying the public or its customers [3]. The latest such date in the data is October 19, 2026 [1].
Every query goes to https://api.fda.gov/device/recall.json with limit=1000 and skip paging, one year at a time, joined with AND to event_date_initiated:[YYYY-01-01 TO YYYY-12-31] [1].
- Software-related: root_cause_description.exact:("Software design" OR "Software Design Change" OR "Software change control" OR "Software Manufacturing/Software Deployment" OR "Software design (manufacturing process)" OR "Software in the Use Environment") OR reason_for_recall:software. These are all six root cause values beginning "Software" [1].
- Cybersecurity, strict: reason_for_recall:(cybersecurity OR cyber OR vulnerability OR vulnerabilities). The query alone matches 40 events and 117 product records [1]. We then read every matched reason and, as Moraph's classification, dropped one 2019 event in which "vulnerability" described an electrical fault, leaving 39 events and 111 product records; the counts for 2020 onward are the same either way [1].
- Cybersecurity, wider: reason_for_recall:(security OR malware OR ransomware OR hacker OR hackers OR CVE OR "unauthorized access" OR malicious OR exploit OR exploited OR "unauthorized user"). This query finds 17 events the strict query missed [1]. Moraph's classification, not FDA's: we kept an event when its reason describes unauthorized access, malware or a weakness in a security control of the device or its data, and dropped it when the word has another sense (a security screw, the security of a tool tip or a suture knot, an incorrect security key that blocked a connection, a Windows security dialog, or a security update after which computers would not restart); by that rule we kept 10 of the 17 and dropped 7 [1].
- Class: records from https://api.fda.gov/device/enforcement.json with recall_initiation_date:[YYYY0101 TO YYYY1231] for each year, matched by event_id; an event is Class I if any of its products is [4].
- Safety communications: the dated rows of the safety communications and alerts table on FDA's medical device cybersecurity page as checked on November 16, 2026, which lists entries from August 17, 2021 to January 30, 2025 and refers readers to FDA's web archive for earlier ones [6]; the earlier rows come from an Internet Archive copy of the page captured January 9, 2024, content current as of December 20, 2023 [7]. FDA does not link communications to recall events. Pairing them is Moraph's match: a communication is paired with a recall event when both name the same manufacturer and the same devices, taking the event closest in time.
Limits: text matching finds only the words firms and FDA chose, so the cybersecurity counts are a floor. Root cause is FDA's own determination and, per FDA policy, can change until the recall is terminated [3]. The posting date in each record is the date FDA classified the recall [3]. By our calculation from the two dates, the median gap from initiation to posting for software-related product records initiated in 2024 and 2025 was 37 days [1]. Our observation, not an FDA statement: recalls begun in September and October 2026 and not yet posted are missing, so the 2026 counts are incomplete and will likely rise. openFDA says its results should be assumed unvalidated and not used for decisions about medical care [1], that FDA may correct earlier recall information, and that recall status is not updated after classification, so we do not use status [2] [5]. Our counts by initiation date differ from counts by enforcement report date, which our 2026 recall data brief uses.
Software recalls: a steady share, a rising severity
| Year initiated | All recall events | Software-related events | Software share | Software events with a Class I product |
|---|---|---|---|---|
| 2016 | 1,232 [1] | 218 [1] | 17.7% [1] | 3 (1.4%) [1] [4] |
| 2017 | 1,091 [1] | 153 [1] | 14.0% [1] | 4 (2.6%) [1] [4] |
| 2018 | 1,074 [1] | 148 [1] | 13.8% [1] | 5 (3.4%) [1] [4] |
| 2019 | 969 [1] | 138 [1] | 14.2% [1] | 10 (7.2%) [1] [4] |
| 2020 | 916 [1] | 141 [1] | 15.4% [1] | 2 (1.4%) [1] [4] |
| 2021 | 900 [1] | 142 [1] | 15.8% [1] | 8 (5.6%) [1] [4] |
| 2022 | 839 [1] | 98 [1] | 11.7% [1] | 6 (6.1%) [1] [4] |
| 2023 | 917 [1] | 122 [1] | 13.3% [1] | 13 (10.7%) [1] [4] |
| 2024 | 973 [1] | 114 [1] | 11.7% [1] | 14 (12.3%) [1] [4] |
| 2025 | 946 [1] | 167 [1] | 17.7% [1] | 17 (10.2%) [1] [4] |
| 2026 (to October 19) | 728 [1] | 104 [1] | 14.3% [1] | 10 of 103 found (9.7%) [1] [4] |
From January 2016 to October 19, 2026 the database lists 10,585 recall events, and 1,545 of them, 14.6%, are software-related by our test [1]. The share has stayed between 11.7% and 17.7% in every year [1]. The yearly count averaged 173 in 2016 to 2018 and 134 in 2023 to 2025, in step with fewer recall events overall, before 167 in 2025, the most since 2016 [1]. Comparing January to August, 2026 had 90 against 105 in 2025, while all recall events rose from 610 to 661 [1].
Severity moved more than volume. In 2016 to 2018, 12 of 519 software-related events, 2.3%, included a Class I product; in 2023 to 2025 it was 44 of 403, or 10.9% [1] [4]. All device recall events moved the same way, from 3.5% to 11.5% [1] [4]. Software recalls share a general rise in Class I classifications rather than leading it.
Moraph's view
Moraph's view: a share near one in seven, year after year, makes software failure a standing feature of device recalls rather than a passing problem. For a hospital, a software recall deserves the same routine as a hardware fault: find the affected units, apply the fix, confirm the installed version.
Cybersecurity in recall records: small numbers, rising since 2023
| Year | Recall events, strict query after reading | Recall events, strict plus wider terms (Moraph's classification) | Entries on FDA's cybersecurity safety communications list |
|---|---|---|---|
| 2016 | 0 [1] | 0 [1] | 0 [6] [7] |
| 2017 | 0 [1] | 2 [1] | 2 [6] [7] |
| 2018 | 2 [1] | 3 [1] | 2 [6] [7] |
| 2019 | 9 (10 matched by the query) [1] | 9 [1] | 3 [6] [7] |
| 2020 | 1 [1] | 1 [1] | 2 [6] [7] |
| 2021 | 2 [1] | 3 [1] | 3 [6] [7] |
| 2022 | 3 [1] | 3 [1] | 3 [6] [7] |
| 2023 | 3 [1] | 3 [1] | 0 [6] [7] |
| 2024 | 5 [1] | 6 [1] | 0 [6] [7] |
| 2025 | 7 [1] | 11 [1] | 1 [6] [7] |
| 2026 | 7 (to October 19) [1] | 8 (to October 19) [1] | 0 (to November 16) [6] [7] |
Recalls that cite cybersecurity are rare: 40 events match the strict query since 2016, 39 after our reading, against 1,545 software-related events [1]. 2019 brought 9, then the count fell to 1 in 2020 before rising to 5 in 2024, 7 in 2025 and 7 in 2026 to October 19, so 22 of the 39 events began in 2023 or later [1]. Adding the wider terms as we classified them gives 49 events, 11 of them in 2025 [1].
Of the 39 strict events, 30 carry one of FDA's software root causes, most often "Software design" with 21 [1]. Only 4 included a Class I product; 32 were Class II, 1 was Class III and 2 were not found in the enforcement records [1] [4]. Recent records name patient and fetal monitors, image viewers and servers, gene sequencing instruments, anesthesia systems, a heart pump controller, a powered wheelchair and a surgical robot console [1]. Several cite third-party components, such as Windows remote desktop in 2019, the Windows Print Spooler and a TCP/IP library in 2021 and an interface engine in 2024, and events in 2025 and 2026 cite exposed login or service credentials [1]. One 2025 record says the flaw was found through CISA's vulnerability disclosure program, and another says the firm acted after receiving an FDA "It Has Come to Our Attention" letter [1].
Safety communications: one entry dated after 2022
FDA's list holds 16 cybersecurity safety communications and alerts dated 2017 to 2026: 15 from 2017 to 2022, none in 2023 or 2024, and one on January 30, 2025, about patient monitors from Contec and Epsimed [6] [7]. As checked on November 16, 2026, the table lists no later entry [6]. The page says that unless noted, FDA is not aware of patient injuries or deaths associated with the alerts listed [6]. Five of the 16 concern components used across many devices, such as BlackBerry QNX, Apache Log4j, the PTC Axeda agent and the Urgent/11 and SweynTooth vulnerability families [6] [7].
By Moraph's match (rule in the method), nine of the 16 pair with a recall event for the same devices in the openFDA data [1] [6] [7]. Firms are named as the recall records list them.
| Safety communication (date and title as FDA lists it) | Recall event initiated, recalling firm as FDA lists it (query that found it) |
|---|---|
| 08/29/2017 [7]: Firmware Update to Address Cybersecurity Vulnerabilities Identified in Abbott's (formerly St. Jude Medical's) Implantable Cardiac Pacemakers | 2017-08-28, St Jude Medical Inc. (wider query) [1] |
| 04/17/2018 [7]: Battery Performance Alert and Cybersecurity Firmware Updates for Certain Abbott (formerly St. Jude Medical) Implantable Cardiac Devices | 2018-04-16, St Jude Medical Inc. (wider query) [1] |
| 10/11/2018 [7]: Cybersecurity Updates Affecting Medtronic Implantable Cardiac Device Programmers | 2018-10-02, Medtronic Inc., Cardiac Rhythm and Heart Failure (CRHF) (strict query) [1] |
| 03/21/2019 [7]: Cybersecurity Vulnerabilities Affecting Medtronic Implantable Cardiac Devices, Programmers, and Home Monitors | 2020-01-30, Medtronic Inc., Cardiac Rhythm and Heart Failure (CRHF) (strict query) [1] |
| 06/27/2019 [7]: Certain Medtronic MiniMed Insulin Pumps Have Potential Cybersecurity Risks | 2019-06-27, Medtronic MiniMed (strict query) [1] |
| 01/23/2020 [7]: Cybersecurity Vulnerabilities in Certain GE Healthcare Clinical Information Central Stations and Telemetry Servers | 2019-11-12, GE Healthcare, LLC (strict query) [1] |
| 06/02/2022 [7]: Illumina Cybersecurity Vulnerability May Present Risks for Patient Results and Customer Networks | 2022-05-03, Illumina, Inc. (strict query) [1] |
| 09/20/2022 [7]: Medtronic MiniMed 600 Series Insulin Pump System Potential Cybersecurity Risk | 2022-09-20, Medtronic MiniMed (strict query) [1] |
| 01/30/2025 [6]: Cybersecurity Vulnerabilities with Certain Patient Monitors from Contec and Epsimed | 2025-04-10, Contec Medical Systems Co., Ltd. (strict query) [1] |
Moraph's view
Moraph's view: the two records now point in opposite directions. Recalls that cite vulnerabilities have risen since 2023 while FDA's own list added one entry in more than four years. Our reading, which FDA has not stated, is that the agency now leaves most public notice of device vulnerabilities to manufacturers' notices and CISA advisories, which its page already links for several entries, and keeps its own safety communications for cases it judges need action by patients or clinicians. A health system that watches only FDA's list will miss most of what the recall records show.
Why most security fixes never become recalls
FDA's final guidance on postmarket management of cybersecurity, issued in December 2016, says the majority of actions manufacturers take to address vulnerabilities, which it calls cybersecurity routine updates and patches, are generally considered device enhancements for which FDA does not require advance notification or reporting under 21 CFR part 806 [10]. Only a small subset of actions that may pose a risk to health must be reported, and the guidance asks manufacturers to assess whether the risk of patient harm is controlled or uncontrolled [10]. A recall that cites a vulnerability therefore marks the cases treated as a reportable correction, not the volume of security patching.
Section 524B of the FD&C Act adds duties at the premarket stage. The sponsor of a submission for a cyber device must submit a plan to monitor, identify and address postmarket vulnerabilities, including coordinated vulnerability disclosure; maintain processes that make updates and patches available; and provide FDA a software bill of materials [9]. The section took effect 90 days after its enactment on December 29, 2022 [9]. FDA's current premarket cybersecurity guidance, issued February 2, 2026, revises the June 27, 2025 version to align with the amended Quality Management System Regulation [6]. For fiscal year 2027, CDRH lists "Postmarket Management of Cybersecurity in Medical Devices" among the draft guidance topics on its B-list, not its A-list [11].
Moraph's view
Moraph's view: the rise in cybersecurity recalls since 2023 fits both the arrival of devices with 524B postmarket plans and more coordinated disclosure, but the numbers are too small to prove either. A revised postmarket guidance could change the count either way by redrawing the line between a routine patch and a reportable correction. Health systems differ most in what they can absorb: a large system with its own clinical engineering and imaging informatics teams can patch a viewer or server in days, while a small hospital often depends on the manufacturer's field service for the same fix.
What to watch
- Whether FDA publishes a draft revision of its 2016 postmarket cybersecurity guidance, now on the CDRH fiscal year 2027 B-list [11].
- Whether the yearly count of strict-query cybersecurity recall events stays at 5 or more, as in 2024, 2025 and 2026 to date [1].
- Whether FDA adds entries to its cybersecurity safety communications list, which has one entry dated after 2022 [6].
What this means for you
- Clinical engineering and biomedical teams: Track software-related recalls with the same workflow as hardware recalls, including confirming the installed software version after the fix; about one recall event in seven since 2016 is software-related [1], and more of them now include Class I products [4].
- CISOs and security teams at health systems: FDA's safety communications list is not a complete feed of device vulnerabilities: it has one entry dated after 2022 [6], while a keyword query of recall reasons matches 22 events since 2023 [1]. Watch recall records, manufacturers' notices and CISA advisories together.
- Supply chain and contracting teams: Ask device makers for the postmarket vulnerability plan and software bill of materials that section 524B requires them to give FDA for cyber devices [9], and for their patch timelines, since most routine security patches are not reported to FDA as recalls [10].
- Imaging, laboratory and IT infrastructure owners: Inventory imaging viewers, archive servers and sequencing instruments with the same rigor as bedside devices: recent cybersecurity recall reasons name these systems and cite exposed credentials and third-party software flaws [1].
Moraph predictions
Every prediction is Moraph's view, not a fact. Each has a date, a confidence level (70% means we expect to be wrong about three times in ten) and a test anyone can apply. We score them on this page when the date comes and never remove one. How we make and score predictions.
P1 By Confidence 75% Open
At least 110 software-related device recall events have an event_date_initiated in calendar year 2027.
How we will check: On or after March 31, 2028, run the software-related query in this report's method against https://api.fda.gov/device/recall.json with event_date_initiated:[2027-01-01 TO 2027-12-31] and count distinct res_event_number values. Right if 110 or more.
Basis: Software-related events numbered 122 in 2023, 114 in 2024, 167 in 2025 and 104 in 2026 to October 19, and the lowest year since 2016 was 98 in 2022 [1].
P2 By Confidence 65% Open
At least 5 device recall events with an event_date_initiated in calendar year 2027 match the strict cybersecurity query and, on reading, concern computer security.
How we will check: On or after March 31, 2028, run reason_for_recall:(cybersecurity OR cyber OR vulnerability OR vulnerabilities) with event_date_initiated:[2027-01-01 TO 2027-12-31] against https://api.fda.gov/device/recall.json, read each event's reason and drop those where the words do not refer to computer security. Right if 5 or more distinct events remain.
Basis: The strict query found 5 events in 2024, 7 in 2025 and 7 in 2026 to October 19, against 1 to 3 a year from 2020 to 2023 [1].
P3 By Confidence 75% Open
FDA's table of safety communications and alerts related to cybersecurity lists at most one entry dated in calendar year 2027.
How we will check: In January 2028, open https://www.fda.gov/medical-devices/medical-device-cybersecurity and count rows in the safety communications and alerts table dated 2027 (or, if the table moves, FDA's successor list). Right if zero or one.
Basis: The list has one entry dated after September 20, 2022, on January 30, 2025, and none in 2023, 2024 or 2026 to November 16 [6] [7].
P4 By Confidence 60% Open
FDA does not publish a draft revision of its guidance Postmarket Management of Cybersecurity in Medical Devices by September 30, 2027.
How we will check: Search FDA's guidance documents database for the title on or after October 1, 2027. Right if no draft or final version dated after December 2016 has been published by September 30, 2027.
Basis: The topic is a draft guidance on the CDRH fiscal year 2027 B-list, not the A-list [11]; the current final guidance was issued in December 2016 [10].
Facts on this page link to their sources. Text marked "Moraph's view", and every prediction, is our own analysis, not a fact or advice. Methodology.
Spotted an error? Tell us and we will check it and correct the piece where needed.