SafelyYou

The model is the engine of two core modules, and the platform carries value beside them. Safety AI detects falls from video in residents' rooms ('Built with learnings from 7M cumulative resident days'), and Staffing AI measures care delivered without wearables, which the company calls 'fully ambient AI'.

Around them sit things that stand without the model: Guardian pendants, buttons and sensors, the Halo emergency call system with virtual check-ins, the Discover portal and reports, and Insight, a remote clinical team of nurses and therapists who review falls with each community.

The oversight structure is described, and the thresholds and limits are not. When a fall is detected, on-site staff are alerted 'within seconds' with clear video of the event; staff decide what to do, and in Halo they can open a virtual check-in to assess urgency before responding. After the event, SafelyYou's remote clinical team assesses each fall against 50 data points and meets the community's staff in fall huddles with nurses, occupational and physical therapists to change care plans.

What is not published: the detection threshold, what the system does not detect or alert on, and what happens when a fall is missed. The company states a false alarm rate ('Only one false alarm per sensor every two years') from its own dashboard, with no method. Monitoring is described as 24/7, which is the scope this structure has to cover.

The approach is described in general terms and nothing is identified. SafelyYou describes AI video fall detection built on 7 million cumulative resident days, ambient care measurement without wearables, and hardware 'designed from the ground up by SafelyYou'. It publishes performance figures from its own internal benchmarking dashboard (2026): on-the-ground incidents detected with '99.25% accuracy', care measured with '99.99% accuracy', and one false alarm per sensor every two years.

Accuracy is a weak measure for a rare event, since a detector that never alerted would also score high; sensitivity (share of real falls caught) and positive predictive value are not published, and the benchmark's method, period and denominator are not described. No model, training data characteristics (lighting, camera placement, resident mobility) or versioning practice is published.

The chain is structurally short at the model and device layers: SafelyYou designs its own Guardian hardware and says its AI is built from its own 7 million cumulative resident days, and fall videos are reviewed by its own remote clinical team, so no outside model provider is described as receiving resident video. The rest of the chain is not enumerated.

The privacy policy (updated 15 July 2026) says information may go to subsidiaries, affiliates, contractors and service providers bound by confidentiality, naming none, and it addresses website, app and text message data rather than resident video. No hosting provider, region or subprocessor list is published.

Published research short of independent validation, and the headline figures rest on internal analytics. The site cites Xiong and colleagues (American Journal of Managed Care, July 2019): a pilot from June to August 2018 in six residential dementia care facilities, 55 residents enrolled, comparing real-time video detection with standard practice.

After 83 falls in the intervention group and 147 in the control group, 15.7 percent against 35.4 percent of falls led to an emergency medical technician visit and 8.3 percent against 24.5 percent to an emergency department visit, relative reductions of 75 and 80 percent. SafelyYou's CEO and COO are coauthors, and the first author is at UC Davis.

The home page's other headline figures (falls down 45 percent in the first year, 82 percent fewer repeat falls within 90 days, 174 days longer stays, 99.25 percent detection accuracy) cite SafelyYou's own 2026 analytics, and the six months longer stay at Milestone Retirement Communities compares residents who opted in with those who did not. The site says there are three peer reviewed studies and 20 case studies; the fall reduction headline has no peer reviewed source named.

General assurances only. The Discover portal where staff watch fall videos is described as 'password protected and maintained with strict security standards', and a 2022 security statement sets out policy principles. Nothing published states when video is captured and when it is not, how long fall videos and virtual check-in footage are kept, who may view them, whether they train the detection models, or whether anything is de-identified; the privacy policy (updated 15 July 2026) covers website, app and text message data, not residents.

The product places cameras in residents' rooms and adds live virtual check-ins in Halo, so these are the questions a family will ask. Safety engineering, meaning how a missed fall or wrong alert is investigated as a safety event, is not published either.

No statement of HIPAA status and no privacy document that reaches resident data. The home, Safety AI, Halo, integrations and security pages make no HIPAA or business associate statement, and the privacy policy (updated 15 July 2026) covers the website, the mobile app and text message consent, with no provision on resident health information or video.

Whether federal health privacy law applies depends on the operator: many assisted living and memory care communities are not covered entities, and where they are not, nothing federal governs video of residents in their rooms.

Controls are asserted with nothing independent behind them. SafelyYou publishes a Security Commitment Statement (updated 14 June 2022) that lists internal policies (information security, network security, physical security, acceptable use) and principles of prevention and incident response, offers single sign-on, and describes its video portal as password protected. No SOC 2 report, HITRUST or ISO/IEC 27001 certification, penetration test or trust center is named. The holding is video of residents in their rooms across more than 1,000 communities.

No clearance is claimed and no regulatory position is stated. The product detects falls and alerts staff, and the site markets fewer falls, fewer fall-related emergency visits, acuity trending and sleep quality assessment. The basis for each fall alert is in front of the staff member who acts on it: clear video of the event, and a clinical review afterward.

The tension is that fall reduction and emergency visit claims are clinical outcome claims, and nothing on the site says how the company classifies the product. Cameras in residents' rooms also raise consent and notice questions under state monitoring and privacy rules, which the company's published material does not address.

Performance is measured and reported only in aggregate, with no framework or results by group. SafelyYou publishes detection accuracy and a false alarm rate from its own internal benchmarking and cites a 2019 peer reviewed pilot, which shows it tests the model; nothing describes a governance process for model changes, and no results are given by lighting, time of night, mobility aid use, posture or skin tone, the conditions under which camera-based detection is known to vary. Falls concentrate at night and among residents with walkers and atypical gait, the people the system exists to protect.

A mechanism lets anyone check an alert, and nothing stands behind the output. Each detected fall comes with video that staff and SafelyYou's clinicians can watch, so a disputed alert can be settled by looking and a missed fall found on review.

The error figures published (99.25 percent accuracy, one false alarm per sensor every two years) come from an internal dashboard with no method or denominator, so they do not stand as an error rate a buyer can test; sensitivity, the share of real falls caught, is not published. The published terms of use cover the website only (as is, liability capped at $100), and no warranty, remediation or route for a resident whose fall went undetected is described.

Named record systems with integration claimed in both directions and little documentation. The home page says SafelyYou is 'available with Yardi, ALIS, PCC, and more', while also saying 'EHR integrations coming soon'; the integrations page shows ALIS, August Health, Yardi and a marketplace partner badge, and says resident profiles and care transitions 'sync automatically between SafelyYou and your EHR', that event data stays accurate in the system of record 'through bidirectional data flows', and that detected falls activate existing nurse call systems with visual context. Single sign-on is offered. No field level documentation, interface specification or export path is published, so which data reaches which system, in which direction, cannot be checked.

Hardware in every room feeding a single hosted service, with location implied rather than committed. Guardian cameras, sensors, pendants and buttons are installed across a community, videos are watched through the Discover web portal, and SafelyYou reports 99.98 percent uptime. No hosting provider, region, tenancy model or on premise option is stated, nothing says whether detection runs on the device or after video leaves the room, and no retention period for fall videos is given.

No price is published, and the shape of the sale can be established: hardware, AI software and remote clinical services per community, sold through a demo. The economic case is published in the operator's terms from SafelyYou's own 2026 analytics: 'Up to $3k' net incremental NOI per resident, 'Up to $100k NOI uplift per community', and 174 days longer stays. The longer stay figure at Milestone Retirement Communities compares residents who opted in with those who did not, which cannot separate the product's effect from who chooses it.

Coverage is named with validation behind part of it. SafelyYou sells to senior living operators and says it 'protects seniors across care levels', with memory care as the center of gravity, across more than 1,000 communities and 50,000 residents. The peer reviewed pilot it cites was run in residential dementia care facilities (AJMC, 2019), so memory care has published validation; assisted and independent living rest on SafelyYou's own analytics and customer quotes. No acute, skilled nursing or in-home care is claimed. The product is sold to facilities, not to home care agencies.