Industries / Construction and property / Commercial and residential construction / Pool construction
Safety
Keeping people from being hurt in and around pools: drains that can hold a swimmer under, alarms and cameras that warn when someone is in trouble, and the chemicals and equipment that can injure.
7
Milestones achieved, from 1997 to 2022.
2
To watch for: announced, not yet confirmed outside.
3
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1997
Dives into shallow home pools
Michael DeVivo and Padmini Sekar of the University of Alabama at Birmingham studied 341 people in the US National Spinal Cord Injury Statistical Center's database, which has enrolled patients since 1973, whose spinal cord injuries came from swimming pool mishaps. Almost all (87%) were hurt in private or residential pools; 57% dove into water less than 4 ft (1.2 m) deep and another 38% into water 4 to 8 ft deep. There were no depth markers in 75% of cases and no warning signs in 87%, and 49% reported drinking alcohol.
Lesson70% of the injuries came from ordinary dives, often on a first visit to that pool, so a pool's depth and markings matter as much as the diver's judgment.
Credit
Michael J. DeVivo and Padmini Sekar, in Spinal Cord
2000
Pool alarms put to test
A pool alarm sounds when something falls into the water. In May 2000 CPSC engineer Troy Whitfield reported tests of four alarms and a wristband at six pools, noting that no voluntary standard yet said how such alarms should perform. Water jugs tied together to the weight of a child of 12 to 15 months (about 18.3 lb) up to one of 42 to 54 months (about 25 lb) were pushed in: alarms mounted on the pool wall below the waterline alarmed more consistently and false-alarmed less than floating surface alarms, and the wristband alarmed in any water, tap water included. ASTM later issued F2208-02, a pool alarm specification; in January 2007 CPSC staff told ASTM their tests found "few products meet the requirements" and that its drop instructions were subjective.
Lessona performance standard is only as good as its test object and how it is dropped; loose wording gives different labs different results.
Credit
Troy W. Whitfield, US Consumer Product Safety Commission (CPSC)
2007
Drain covers become law
The suction at a pool or spa drain can hold a swimmer underwater. In 2002 Virginia Graeme Baker, 7, a granddaughter of former US Secretary of State James Baker, died after sitting on the floor drain of a hot tub; CPSC data show 83 suction entrapments from 1999 to 2008, with 11 deaths and 69 injuries. The Virginia Graeme Baker Pool and Spa Safety Act, signed on 19 December 2007, required from a year later that every drain cover sold in the US meet the ASME/ANSI A112.19.8 standard and that every public pool and spa use such covers. A public pool with a single main drain that a body could block also needs a backup, such as a safety vacuum release system that stops or reverses the pump when the drain is blocked. CPSC's 2019 entrapment report says that since the Act took effect in December 2008 no child has been reported killed by entrapment on a drain cover in a public pool or spa; it counted 11 entrapment victims from 2014 to 2018, 2 of them killed, both in spas.
Credit
Nancy Baker and other families who lost children; the US Congress; CPSC, which enforces it
2011
A cover is not enough
In 2010 CPSC ruled that a large enough drain cover could make a small single drain count as "unblockable", so the pool needed no backup system; ABC News reported this could affect as many as 150,000 public and hotel pools and hot tubs. After 156 letters arguing that covers of any size come off or break during a pool's life, the commission voted 3 to 2 to revoke the ruling, effective 11 October 2011: "Regardless of the size of a drain and its cover, the drain cover can come off, presenting a risk of entrapment."
Lessona single part, however well made, is not a layer of protection on its own; the law's intent was layers.
Credit
CPSC Chairman Inez Tenenbaum and Commissioners Robert Adler and Thomas Moore, who voted for it; the families who wrote in
2017
Standards for drowning cameras
A computer vision drowning detection system watches a pool through cameras and alerts staff when a swimmer may be drowning. ISO 20380:2017 sets safety requirements and test methods for such systems in public pools; Britain adopted it as BS EN ISO 20380:2017 on 31 May 2018. For home pools, ASTM F3698-24, last updated in May 2024, requires a system to signal within 30 s that a person is drowning or behaving as a drowning person does, and to tell users when the water is too murky to see.
Credit
ISO; ASTM Subcommittee F15.49
2018
Feeders run without flow
Public pools dose chlorine and acid through automatic feeders that should stop when the circulation pump stops. In August 2018, after a power cut at an outdoor pool in upstate New York, the flow switch that should have shut the feeders failed; chlorine and acid mixed in the pipes, and when the pump was restarted chlorine gas reached the pool, giving people blisters, nausea and eye irritation. CDC's count from emergency room records shows an estimated 13,508 visits for pool chemical injuries in the US in 2015–2017, 36.4% of them children, and at least 56.3% at a home.
Lessona required interlock still has to be tested; the pool stayed shut for the season, the flow sensor was replaced and checks every 2 hours were added.
Credit
Kayla Vanden Esschert, Joseph Laco and colleagues at CDC, with the New York State Department of Health
2022
Headband alarms tried at camp
A wearable drowning alarm signals staff when the swimmer wearing it stays underwater too long. Molly Johnson and Karla Lawson ran the WAVE system, headbands that alert staff bracelets after more than 20 s underwater, for 120 hours at a YMCA camp pool in central Texas in summer 2021, with 120 to 180 children a day. It gave one or two alerts an hour, mostly for risky underwater play or children leaving the pool area still wearing a headband; 93% of staff agreed it could help save a life, and 66% disagreed that children did not mind the headbands. The authors note there is "almost no research" on whether water safety devices prevent drowning, apart from pool fencing and life jackets.
Lessonwith no drowning in a season, a field trial can measure alerts and acceptance, not lives saved.
Credit
Molly B. Johnson and Karla A. Lawson, Dell Children's Medical Center, Austin
Watch for
Cameras that alert lifeguards
Camera systems that alert lifeguards through a smartwatch are spreading in public pools. The City of Newcastle, Australia, said its trial of Lynxight cameras at Wallsend Swimming Centre would help lifeguards respond "up to six times faster"; in August 2026 Seven's Sunrise, reported by The Nightly, said the system was being rolled out to about 100 Australian facilities and had "reportedly already saved two lives". In July 2026 the YMCA of Middle Tennessee installed AngelEye, which pairs ceiling cameras with sensors in the underwater lights.
Achieved whenan outside evaluation publishes the detection rate, false alarms per hour and lifeguard response times for pools with and without the system, for example using the tests in ISO 20380.
Claimed by
City of Newcastle, November 2025; The Nightly and Sunrise, August 2026
One wire or a grid
Equipotential bonding ties a pool's steel, metal parts and the deck around it together, so that a fault cannot set up a dangerous voltage between surfaces a swimmer touches. The National Fire Protection Association, which writes the US National Electrical Code (NEC), published a tentative interim amendment saying only copper or steel grids should be used around the pool edge; the trade's most common method is a single loop of #8 AWG copper wire. A study commissioned by the Pool Industry Council, the Florida Swimming Pool Association's research nonprofit, with funding and technical guidance from the Pool & Hot Tub Alliance, measured nine new pools in Central Florida over nine and a half weeks. It found that grids did slightly better but that both methods kept the modeled current through a person, taken as 500 ohms, well below 2 mA under simulated faults of up to 120 V.
Achieved whena lab or research group with no stake in the code change repeats the field measurements on more pools and shell types and publishes the data.
Claimed by
SunSmart Engineering for the Pool Industry Council, August 2025
Open
01
Pool alarms tested again
A repeatable public test of the pool alarms sold today, floating, wall-mounted, wearable and camera types alike, using the same weighted test objects and the same drop for each, plus a month of logs for false alarms in wind and rain.
The briefCredit · unclaimed02
One test for drowning cameras
An openly licensed set of videos from real pools in which weighted manikins sink among normal swimmers, labeled to the second, with a scoring script, so research models and commercial systems can be compared on the same events: share detected within 30 s, time to alert and false alarms per hour.
The briefCredit · unclaimed03
Entrapments counted since 2018
An open table of every suction entrapment reported in US pools and spas since 2019, each with the date, pool type, public or home, drain and cover type, whether a backup system was fitted, and the outcome, set beside CPSC's earlier counts.
The briefCredit · unclaimed
Open 01 · the brief
Pool alarms tested again
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A repeatable public test of the pool alarms sold today, floating, wall-mounted, wearable and camera types alike, using the same weighted test objects and the same drop for each, plus a month of logs for false alarms in wind and rain.
Why it is still open
The most recent government test found is CPSC's of May 2000, of five products. In 2007 CPSC staff told ASTM that few products met its standard and that the drop instructions were subjective. Tests since are informal: Bob Vila's reviewers tested nine alarms over about 6 hours for sensitivity, range and loudness, with no standard test object, and the Dell Children's researchers found almost no research on whether such devices work.
Built with what exists
A home pool used with the owner's permission, with nobody in the water; test objects of water jugs tied to 18 lb and 25 lb, as in CPSC's 2000 test; a drop rig that releases the object the same way every time; phone video at 240 frames a second to time each alarm; a sound level meter; a small weather station; Python.
First step
Build the drop rig and test objects and run 20 drops each on one floating and one wall-mounted alarm.
Done looks like
At least 10 alarms, 50 drops each at two weights and three distances, with the detection rate and time to alarm for each; false alarms per day over 30 days; results and video published.
Open 02 · the brief
One test for drowning cameras
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An openly licensed set of videos from real pools in which weighted manikins sink among normal swimmers, labeled to the second, with a scoring script, so research models and commercial systems can be compared on the same events: share detected within 30 s, time to alert and false alarms per hour.
Why it is still open
Mercy Chepkorir and colleagues reviewed 30 machine-learning drowning detection studies in November 2026: reported accuracy averaged above 90.9%, but only 1 study was validated under real operating conditions, none was tested on a second dataset, and the field lacks standard datasets. The closest benchmark found, SwimID from Northwest A&F University in China, uses pool CCTV footage of struggling, floating and sinking but can only be had by signing a release agreement, for academic use. ISO 20380 and ASTM F3698 set test methods, but no published results by product were found.
Built with what exists
Water rescue training manikins weighted to sink; closed pool sessions with the operator's permission; ceiling and underwater cameras such as GoPros; lifeguards who swim normally as background and never act out distress, filmed with their written consent; CVAT for labeling; Ultralytics YOLO for a baseline detector; Hugging Face to host the set; Python for scoring.
First step
Film 20 manikin submersions and 2 hours of normal lane swimming at one pool in a closed session, and label them.
Done looks like
500 staged manikin events and 50 hours of event-free swimming across 5 pools and 3 lighting conditions; a baseline model and at least one commercial system scored by the same script, with results published.
Open 03 · the brief
Entrapments counted since 2018
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open table of every suction entrapment reported in US pools and spas since 2019, each with the date, pool type, public or home, drain and cover type, whether a backup system was fitted, and the outcome, set beside CPSC's earlier counts.
Why it is still open
CPSC's 2026 drowning report leaves entrapments out and points to a separate report for them, covering 2014 to 2018 and published in May 2019, which counted 11 victims; no later count was found. Raw material exists: a Hugging Face set of 18,464 drowning news articles from 2019 to July 2026, which measures news coverage, not deaths.
Built with what exists
News archives and that news set; court records; CPSC's SaferProducts.gov reports through its public API; NEISS emergency room records; a web agent such as Meta's Muse Ember (see the Note "The jobs agents repeat") to gather reports; a model that answers fixed questions about each report (see the Note "A thousand questions, one pass"), with a person checking each case.
First step
Collect the 2019 and 2020 cases from news archives and code the first 20 by hand.
Done looks like
Every case from 2019 to 2025 with at least one source, checked by two people, and counts by year and by public or home pool set beside CPSC's 83 reports from 1999 to 2008.