Industries / Consumer goods / Baby products
Monitors
Baby monitors let a parent hear, see or check on a sleeping baby from another room: radio and video units, movement pads under the mattress, socks and bands that read pulse or oxygen, cameras and radar that watch the chest move, and the apps and cloud servers behind them.
7
Milestones achieved, from 1937 to 2024.
3
To watch for: announced, not yet confirmed outside.
3
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1937
A radio for the nursery
Eugene F. McDonald Jr., president of the Zenith Radio Corporation of Chicago, wanted to listen in on his young daughter, and Zenith had the sculptor Isamu Noguchi design the Radio Nurse: a transmitter called the Guardian Ear, placed by the crib, and a Bakelite receiver for the parents, which carried sound over the house wiring. The Victoria and Albert Museum dates the design to 1937 and calls it one of the earliest baby monitors to be made commercially; The Henry Ford museum says Zenith put it on sale in March 1938 at USD 19.95. Both museums tie the worry behind it to the 1932 kidnapping and murder of the aviator Charles Lindbergh's baby son.
Credit
Isamu Noguchi; Eugene F. McDonald Jr. and Zenith Radio Corporation
1972
Apnea blamed for crib death
Sudden infant death syndrome (SIDS) is the unexplained death of a baby, usually in sleep. In Pediatrics in October 1972, Alfred Steinschneider, a doctor in Syracuse, New York, described two babies of a mother who had already lost three children; both showed long pauses in breathing (apnea) in his sleep laboratory and died soon after going home. He argued that apnea was "part of the final pathway" to SIDS and could run in families. Home apnea monitors, which sound an alarm when breathing stops for a set time, spread in the years that followed; the science writer Sharon Begley wrote that the paper "gave birth to a multimillion-dollar apnea-monitor industry."
Lessonin 1995 the mother, Waneta Hoyt, was convicted of murdering all five of her children, including the two in the paper, so the cases that launched home apnea monitoring were not SIDS at all.
eutils.ncbi.nlm.nih.govdiscovermagazine.comencyclopedia.comsharonlbegley.com
Credit
Alfred Steinschneider
1999
Pads under the mattress
A movement monitor sounds an alarm when it stops sensing a baby's movement for a set time, usually 20 seconds by the UK charity the Lullaby Trust's account. Angelcare, of Quebec, sold monitors with a sensor pad under the crib mattress from October 1999; its patent, filed in July 1998 by Maurice Pinsonneault and John Keith Millns, describes a piezoelectric transducer, a crystal that gives off a voltage when pressed, held between two plates. The Care of Next Infant (CONI) programme in England, Wales and Northern Ireland still offers such pads to parents whose earlier baby died suddenly.
Lessonin November 2013 Angelcare and the CPSC recalled about 600,000 of the monitors for repair after two girls, aged 8 months and 13 months, died when they pulled the pad's cord, about 11 feet long, into the crib, so the part that made the alarm work had become the hazard.
Credit
Angelcare Monitors Inc.; Maurice Pinsonneault and John Keith Millns
2001
Alarms are not warnings
The Collaborative Home Infant Monitoring Evaluation (CHIME) followed 1,079 babies, healthy term babies and groups at higher risk of SIDS, at five US medical centers from 1994 to 1998, over 718,358 hours of home monitoring. Events that crossed the usual alarm limits happened in 41% of the babies and were "quite common, even in healthy term infants"; extreme events happened in 10%, mostly in babies born early, and the authors found them "not likely to be immediate precursors to SIDS." In 2003 the American Academy of Pediatrics (AAP) wrote that "multiple studies have been unable to establish" that home monitoring works, and its 2022 safe sleep advice says not to use home monitors of breathing and heartbeat as a way to lower the risk of SIDS.
Lessonbecause SIDS is rare, large studies of whether monitors prevent deaths have never been run, by the Lullaby Trust's account, so a monitor can reassure a family without anyone knowing whether it saves lives.
eutils.ncbi.nlm.nih.goveutils.ncbi.nlm.nih.govhealthychildren.orglullabytrust.org.uk
Credit
R. Ramanathan, M. J. Corwin, C. E. Hunt and the CHIME Study Group; AAP
2021
A sock ruled a device
Owlet's Smart Sock reads a baby's pulse and blood oxygen at the foot by pulse oximetry, shining light through the skin. In August 2018 Christopher Bonafide and colleagues at the Children's Hospital of Philadelphia compared the Owlet Smart Sock 2 and the Baby Vida with a hospital oximeter on 30 babies up to six months old on hospital wards: for low oxygen (below 91%), the Owlet caught 88.8% of true low readings and the Baby Vida none, and the Baby Vida showed falsely low pulse rates in 14 babies. On 1 October 2021 the US Food and Drug Administration (FDA) sent Owlet a warning letter: socks meant to detect low oxygen and slow heartbeat and to alarm when readings leave preset values are medical devices, Owlet was selling them without clearance, and the FDA had told the company so since 2016. In December 2021 Owlet said it would stop selling the Smart Sock and seek FDA authorization.
Lessona product sold as reassurance becomes a medical device once it claims to catch a baby's low oxygen or slow heartbeat, and must then show that it does.
Credit
C. P. Bonafide, A. R. Localio, D. F. Ferro and colleagues (JAMA, 2018); FDA
2023
A class for home oximeters
On 8 November 2023 the FDA granted Owlet's De Novo request, the route for a new type of device with no existing equivalent, for the Dream Sock, and so created a class called "infant pulse rate and oxygen saturation monitor for over-the-counter use". Its special controls require clinical testing of the effect of skin pigmentation on readings, bench testing that the device recognizes a poor signal and that its alarms work, and usability testing showing that caregivers read the outputs correctly. Masimo's Stork, which reads oxygen, pulse rate and skin temperature in babies from birth to 18 months, was cleared into the same class on 3 May 2024, and the FDA wrote the class into the Code of Federal Regulations (21 CFR 870.2705) on 9 October 2026.
Lessonin September 2025 the FDA told caregivers not to use unauthorized infant monitors that claim to measure vital signs, made by dozens of companies, and stated that "No devices, including infant monitors, have been authorized to prevent SIDS or SUID" (sudden unexpected infant death).
accessdata.fda.govpublic-inspection.federalregister.govfda.innolitics.comsleepreviewmag.comfda.gov
Credit
FDA Center for Devices and Radiological Health; Owlet Baby Care; Masimo
2024
Security rules reach monitors
In September 2013 the US Federal Trade Commission (FTC) settled with TRENDnet, whose SecurView cameras were marketed for uses including baby monitoring, after hackers posted links to the live feeds of nearly 700 of them, one showing babies asleep in their cribs. In 2015 the security firm Rapid7 graded nine internet-connected baby monitors from eight vendors on a 250-point scale: eight got an F and one a D, with hidden hardcoded passwords and unencrypted video among the flaws. From 29 April 2024 the UK's Product Security and Telecommunications Infrastructure (PSTI) regime banned universal and easily guessed default passwords on consumer connected products, and required makers to publish a way to report security bugs and the minimum period they will supply security updates. The EU's Cyber Resilience Act, in force from 10 December 2024 and applying from 11 December 2027, lists "baby monitoring systems" among its class I products and requires makers to give an early warning within 24 hours of learning that a flaw is being exploited.
Lessonin January 2023, before either law applied, the consumer testers Test-Achats (Belgium) and Deco Proteste (Portugal) found 69 vulnerabilities across 17 connected products for babies and children, five with a critical flaw.
ftc.govrapid7.comgov.ukspringlex.euoeil.europarl.europa.eueuroconsumers.org
Credit
FTC; Rapid7; UK government; European Parliament and Council of the EU
Watch for
Radar breathing monitors
Momcozy says its BM08, launched in May 2026 at USD 249.99, is "The First Contact-Free Baby Monitor to Track Breathing and Heart Dynamics Using Radar Technology." It uses millimetre-wave radar, radio waves a few millimetres long that bounce off the chest and register its rise and fall without touching the baby. The release gives no accuracy figures and no medical device status, and the FDA's over-the-counter infant monitor class lists only two devices, both pulse oximeters.
Achieved whenan independent tester or a peer-reviewed study reports how often and how fast it alarms when breathing stops, or a regulator authorizes it for infants.
Claimed by
Momcozy, May 2026
Predicting development from video
Nanit, whose crib camera reads breathing from the movement of a patterned band, raised USD 50 million, led by Springcoast Partners, for a "Parenting Intelligence System" set to launch in 2026. It is to track movement, breathing, motor milestones and speech, and look for trends that may predict "metabolic, emotional, or cognitive challenges." The company says babies on its current system "sleep 10% longer on average" and that it holds over 5 billion hours of infant sleep data.
Achieved whenresearchers independent of Nanit publish a study showing that the system's early flags match later clinical assessments, with its false-alarm rate reported.
Claimed by
Nanit, reported January 2026
AI sleep coaching
Owlet's chief executive, Jonathan Harris, told investors on 5 March 2026 that the company is "evolving from simple hardware monitoring to real-time personalized AI sleep coaching", with an AI Sleep Insights feature to turn its data into daily plans for parents, and reported more than 110,000 paying subscribers. No launch date or test results were given.
Achieved whena randomized trial run or analyzed by outside researchers shows that families using the coaching get measurably better infant sleep or parent well-being than families using the same monitor without it.
Claimed by
Owlet, March 2026
Open
01
Breathing manikin bench test
A public bench test of camera and radar baby monitors that claim to watch breathing, using an infant CPR training manikin whose chest is moved by a servo-driven plate at set rates and depths. The rig stops the motion on a schedule, and one camera films the manikin, an LED that lights when the motion stops, and the parent's phone together, so the delay from stop to alarm can be read frame by frame. No child takes part.
The briefCredit · unclaimed02
Security audit against the baseline
A public audit of at least 20 Wi-Fi baby monitors on sale in the UK and EU, scoring each against the three UK PSTI requirements and those provisions of ETSI EN 303 645, the European baseline standard for consumer connected devices, that can be checked from outside the device (default passwords, update policy, encryption of video and logins, what data leaves the home). It also maps every server each camera and app contacts during 24 hours of use with a doll in the crib. Testers watch only the traffic of devices and accounts they own, do not probe the makers' servers, and send any flaw to the maker under coordinated disclosure before publishing.
The briefCredit · unclaimed03
Cry alerts under household noise
A test of whether monitors with cry detection or "cry translation" raise an alert when a recorded cry plays from a speaker beside a doll in a crib, with household noise (a white-noise machine, television speech, a vacuum cleaner, a dishwasher) mixed in at set loudness levels, and how often noise alone sets them off. For translation features, the same clip is played many times to see whether the label stays the same. Recordings come only from public, licensed sound libraries.
The briefCredit · unclaimed
Open 01 · the brief
Breathing manikin bench test
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public bench test of camera and radar baby monitors that claim to watch breathing, using an infant CPR training manikin whose chest is moved by a servo-driven plate at set rates and depths. The rig stops the motion on a schedule, and one camera films the manikin, an LED that lights when the motion stops, and the parent's phone together, so the delay from stop to alarm can be read frame by frame. No child takes part.
Why it is still open
The only manikin bench test of an infant breathing sensor found is a January 2026 study of PneumoWave, a single chest-worn motion sensor: three manikins were driven by ventilators at 6 to 80 breaths a minute with pauses of 5 to 20 seconds, and the pauses were marked by hand afterwards rather than timed as alarms. Reviewers at BabyGearLab tried a doll on Nanit and "had trouble tricking the device into thinking our fake baby had stopped breathing," but published no measurements. The Lullaby Trust says it does not "carry out any testing or research on products." The FDA's over-the-counter infant monitor class covers only devices that shine light through the skin, and no camera or radar monitor is in it.
strathprints.strath.ac.ukbabygearlab.comlullabytrust.org.ukpublic-inspection.federalregister.gov
Built with what exists
An infant CPR manikin such as Laerdal's Resusci Baby or PractiMan's PractiBaby (both used in the 2026 study); a hobby servo or small linear actuator driven by an Arduino or Raspberry Pi Pico; a low-power heating pad so presence sensors see a warm body; a phone camera; Python and OpenCV to read delays from the video; Claude Fable 5.2 to help draft the controller board (see the Note "Three working devices").
First step
Buy one infant CPR manikin and one camera monitor with a breathing feature, build the servo chest plate, and run 20 timed stops to check that the monitor accepts the manikin at all.
Done looks like
At least four monitors (two camera-based, two radar), each put through 50 breathing stops of 30 seconds at three breathing rates, in light, in darkness and with the manikin under a blanket; for each monitor, the share of stops that raised an alarm and the median and longest delay, false alarms over 24 hours of uninterrupted simulated breathing, and a note of any monitor that would not accept the manikin as a baby; videos, code and data published.
Open 02 · the brief
Security audit against the baseline
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public audit of at least 20 Wi-Fi baby monitors on sale in the UK and EU, scoring each against the three UK PSTI requirements and those provisions of ETSI EN 303 645, the European baseline standard for consumer connected devices, that can be checked from outside the device (default passwords, update policy, encryption of video and logins, what data leaves the home). It also maps every server each camera and app contacts during 24 hours of use with a doll in the crib. Testers watch only the traffic of devices and accounts they own, do not probe the makers' servers, and send any flaw to the maker under coordinated disclosure before publishing.
Why it is still open
Earlier tests are old, partial or anonymous. Rapid7 graded nine monitors in 2015; Which? found a critical flaw in one baby monitor app service in June 2021; the 2023 Euroconsumers tests of 17 products named no brands. In May 2026 an independent researcher, Sammy Azdoufal, showed that the alert service of Meari, a camera platform he says serves more than 300 brands and 1.1 million registered devices, let any logged-in account receive messages from devices it did not own (CVE-2026-33356, coordinated by runZero); because the fault was in the cloud, he wrote, a firmware update on the camera would not fix it. The UK rules have applied since April 2024 and are aligned with EN 303 645, but no published audit of current baby monitors against them was found.
rapid7.comwhich.co.ukeuroconsumers.orggithub.comrunzero.comgov.uk
Built with what exists
A Raspberry Pi set up as a Wi-Fi access point; Wireshark and mitmproxy to record and inspect the devices' own traffic; nmap to list the devices' open network ports; the makers' own phone apps; Browser Use 1.0 to collect each maker's security policy and support pages (see the Note "Deleting 80% to start over"); Gemini Decision 1 to score the yes-or-no checklist items across those pages (see the Note "A thousand questions, one pass"), checked by hand on a sample.
First step
Buy the three cheapest Wi-Fi baby monitors on a UK marketplace, record 24 hours of their traffic through a Raspberry Pi access point, and check whether each maker publishes a bug-report contact and a support period as the UK rules require.
Done looks like
20 monitors from at least 10 brands, each with a scored checklist, a list of every domain and country contacted, and whether video and logins travel encrypted; two testers score five devices independently and agree on at least 90% of items; all findings sent to makers at least 90 days before publication.
Open 03 · the brief
Cry alerts under household noise
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A test of whether monitors with cry detection or "cry translation" raise an alert when a recorded cry plays from a speaker beside a doll in a crib, with household noise (a white-noise machine, television speech, a vacuum cleaner, a dishwasher) mixed in at set loudness levels, and how often noise alone sets them off. For translation features, the same clip is played many times to see whether the label stays the same. Recordings come only from public, licensed sound libraries.
Why it is still open
The accuracy figures are the makers' own. Maxi-Cosi said in January 2024 that the cry feature in its See Pro 360 monitor, from the company Zoundream, has "cry detection accuracy just under 90% and cry translation accuracy above 90%"; Philips, which uses the same service, tells users to keep the room free of background noise "such as pets, dehumidifiers, or vacuum cleaners." In research, a cry detector that scored an F1 of 0.656 in the lab (F1 combines missed cries and false alarms; 1 is perfect) scored 0.236 on real home audio, by the count of Xuewen Yao and colleagues, and a 2025 paper tests a research model at several noise levels but no commercial monitor. BabyGearLab's sound tests measure how loud the parent unit plays, not whether cries are detected. No independent test of commercial cry alerts was found.
notebookcheck.netusa.philips.comarxiv.orgarxiv.orgbabygearlab.com
Built with what exists
A powered speaker and a sound level meter; a Raspberry Pi to play cries and noise on a logged schedule; licensed clips from public sound libraries such as Freesound, with each clip's licence checked; phones running the monitor apps, with screen recording; Python to match alerts to plays; a small open audio model fine-tuned on one graphics chip, such as one of Liquid AI's LFM3 models, as a baseline detector (see the Note "One GPU to fine-tune").
First step
Choose 20 licensed cry clips and four noise tracks, write the playback schedule, and run it against one monitor at three noise levels.
Done looks like
At least five monitors; 200 cry plays each at four noise levels and two distances; the share detected and the median delay; false alerts per hour over 10 hours of noise with no cries; for translation features, the share of repeated plays given the same label; clip list, schedule and logs published.