For generations, vehicle safety engineers have learned from collisions by fastening instrumented dummies into cars and crashing those cars under tightly controlled conditions. Toyota is now helping push that process further into the digital world. On September 8, 2026, safety supplier Autoliv announced Toyota as the first automaker to begin evaluating its new Human Body Model Safety Suite.
The development does not mean physical crash testing is disappearing. Instead, detailed computer models of bones, muscles, organs and other anatomy are increasingly being used alongside conventional tests. For Toyota, which has spent decades developing its own virtual-human technology, the Autoliv evaluation reflects a broader industry shift: engineers increasingly want to understand not simply how a dummy moves during an impact, but what may actually happen inside a human body.
Toyota Is the First Customer for Autoliv’s New Safety Platform
Autoliv describes Toyota as the first customer onboarded to evaluate its Human Body Model, or HBM, Safety Suite. Toyota has recently started using the platform as part of its virtual safety-development activities. The system combines the SAFER Human Body Model with visualization and analysis software intended to turn enormous amounts of crash-simulation data into information engineers can use when designing restraints, seats and vehicle structures.
That distinction matters. Toyota is not discovering digital crash testing for the first time. Rather, it is becoming the first automaker to evaluate this particular integrated Autoliv platform. Autoliv says its suite can handle applications involving vehicle occupants as well as pedestrians, cyclists and motorcyclists. The supplier plans broader industry availability during 2026. For engineers accustomed to watching slow-motion footage of a physical crash, the new approach offers another perspective: a collision can be examined virtually from inside the chest, spine or other body regions before a prototype vehicle is sacrificed in a test laboratory.
A Virtual Human Can Reveal What a Dummy Cannot
A traditional crash dummy is an extraordinarily sophisticated measuring instrument, but it remains an engineered surrogate. Its sensors can record acceleration, forces, moments and deformation at selected locations. Human Body Models take a different approach. They digitally represent anatomy and use finite-element calculations to estimate how different structures of the body move, deform and experience loading during an impact.
Autoliv says its model can provide information about occupant response and possible injury mechanisms that physical dummies alone cannot supply. Toyota makes a similar argument about its own THUMS technology, which can represent structures including the skeleton, brain, internal organs and muscles. That does not make a virtual human inherently more trustworthy than a well-established physical test. Each tool has different strengths. NHTSA consequently researches both advanced crash-test dummies and human body models. The emerging strategy is therefore less about choosing digital humans over dummies than combining physical measurements, computational biomechanics and real-world crash evidence to expose weaknesses that one testing method might miss.
Toyota Has Been Building Digital Humans Since the 1990s
Toyota’s involvement in virtual crash testing stretches back nearly three decades. The company began developing its Total Human Model for Safety, better known as THUMS, with Toyota Central R&D Labs in 1997. Version 1 arrived in 2000, which Toyota described as the world’s first virtual human body model software capable of simulating and analyzing whole-body injuries from vehicle collisions.
The technology became progressively more detailed. Later generations added more precise models of the face, brain and internal organs, followed by different body sizes, children and muscular activity. Toyota announced in 2020 that THUMS would become freely available from January 2021; at that point, it said the technology was already being used by more than 100 automakers, suppliers, universities and research institutions. That history makes Toyota’s decision to evaluate Autoliv’s system especially notable. It is not an automaker abandoning an older method for something unfamiliar. It is an experienced developer of human modeling examining another platform as virtual safety engineering becomes more collaborative and sophisticated.
The Hard Part Is Proving the Digital Body Behaves Like a Real One
A convincing-looking computer body is not automatically a reliable injury-prediction tool. Engineers must validate how the model behaves against biomechanical evidence. Toyota says THUMS has been validated using component and whole-body loading tests reported in scientific literature, including 38 tests involving postmortem human subjects. Such comparisons help determine whether simulated bones, tissues and body movements respond realistically when subjected to crash-like forces.
The SAFER model behind Autoliv’s system has undergone similar research. A 2026 study described a validation framework for predicting rib-fracture risk with SAFER HBM V11.1.0. Researchers reconstructed frontal tests involving 43 postmortem human subjects and additional oblique and lateral datasets. For frontal impacts, the model achieved an area-under-the-curve value of 0.90 when predicting the risk of two or more fractured ribs. Results such as these illustrate why validation matters. A useful digital human cannot simply produce detailed graphics; its predicted injury patterns need to correspond closely enough with experimental evidence to earn engineers’ and safety organizations’ confidence.
Digital Models Could Help Safety Testing Represent More People
One attraction of virtual testing is the potential to examine occupants who cannot all be represented economically by fleets of physical dummies. Human beings differ in sex, age, body mass, anatomy, bone strength and posture. Toyota’s current THUMS lineup includes multiple male and female body sizes, child models and specialized representations, while NHTSA supports development of human body models intended to investigate differences in occupant injury risk.
Those differences have measurable consequences. NHTSA found that, among comparable occupants in model-year 2010–2020 vehicles, estimated female fatality risk remained 6.3% higher than male risk, although the gap was dramatically smaller than in much older vehicles. The agency is also developing advanced physical female dummies, including THOR-05F, making clear that diversity in virtual models and better physical surrogates are complementary efforts. Computational models can make it easier to explore numerous body characteristics or seating conditions before deciding which designs deserve expensive physical validation. In practical terms, the future crash lab may test not one standardized occupant, but a digital population before the first real vehicle hits the barrier.
Safety Ratings Are Beginning to Make Room for Virtual Humans
Virtual crash simulations have been used inside automakers for years, but incorporating them into formal consumer safety assessments requires a much higher level of standardization. Euro NCAP has been developing that framework. Its Vision 2030 strategy calls for virtual testing to be progressively integrated into its safety program, while technical procedures now specify how human body models must be qualified before they can be used in prescribed virtual crash scenarios.
The transition is deliberately cautious. Euro NCAP’s HBM qualification protocol began implementation on January 1, 2026, with initial frontal HBM results used for monitoring rather than injury scoring. The organization says strain-based evaluation of rib-fracture risk is planned from 2029. Qualification requirements cover factors such as anthropometry, model quality, validation and standardized outputs. That process shows why platforms such as Autoliv’s are strategically important. Once virtual results influence formal vehicle ratings, automakers will need repeatable simulations that regulators and independent assessment organizations can trust—not simply proprietary computer models that produce impressive internal engineering presentations.
Physical Crash Dummies Are Not Heading for Retirement
It would be tempting to imagine a future in which every crash happens on a computer and pristine prototypes never meet concrete barriers. Current safety practice points somewhere else. Autoliv explicitly describes virtual testing as a complement to physical crash testing, and regulators continue maintaining extensive physical testing programs. NHTSA, for example, operates a dummy-management laboratory supporting regulatory, consumer-testing and research programs while simultaneously developing finite-element human models.
Physical tests provide something simulation cannot generate independently: direct evidence of how an actual vehicle, seat, belt, airbag and structure behave under violent real-world forces. Computer models depend on assumptions and validated representations of those physical systems. Physical testing, meanwhile, cannot practically examine every combination of occupant, impact angle, seating posture and restraint setting. Combining the two therefore creates a feedback loop. Simulations can reveal interesting or troublesome scenarios; physical experiments can verify the models; improved models can then investigate many more variations. The crash dummy remains important, but it increasingly shares the laboratory with its digital counterpart.
Virtual Testing Can Move Safety Decisions Earlier in Development
Building a vehicle specifically to destroy it is expensive and time-consuming. Even sled tests require hardware, preparation and instrumentation. Toyota has said computer simulations allow engineers to repeat many collision patterns while reducing development lead times and costs. Autoliv similarly argues that virtual testing can give engineers useful information earlier, before a vehicle program reaches the stage where complete prototypes are readily available.
That can alter the way safety problems are solved. Imagine an engineer considering several seat-belt geometries, airbag strategies or interior structures. Physical testing might narrow the practical number of configurations that can be explored. A virtual environment can screen far more variations, identify promising designs and send the most important cases forward for hardware testing. It can also investigate complex conditions that are difficult to reproduce repeatedly. Autoliv says its new suite is intended to translate complex simulation output into clearer information for decision-making. The largest benefit may therefore be less dramatic than eliminating crash tests: engineers can potentially make better choices earlier, when changing a component is easier than redesigning it late in development.
Reclining Seats and Pre-Crash Braking Create New Safety Questions
Future vehicle interiors make virtual humans especially useful because occupants may not always sit bolt upright in familiar positions. NHTSA is researching protection for unconventional configurations expected in vehicles with automated-driving systems, including forward-facing and rear-facing reclined seats. Toyota’s current THUMS lineup likewise includes reclined occupant models at several seat-back angles, reflecting the growing importance of posture in crash biomechanics.
Research using the Active SAFER Human Body Model demonstrates how complicated those scenarios can become. A 2024 study modeled a person reclined at 50 degrees, subjected first to automated emergency braking and then to a 50 km/h frontal crash. Researchers compared conventional B-pillar-mounted belts with belt-in-seat configurations and found substantial differences in body movement and predicted injury risks. The important point is not that one simulation determines how every future seat should be designed. It is that occupant posture can change before impact, and muscles, restraints and seating geometry interact across the entire event. A rigid dummy in one standardized position can answer only part of that problem.
The Bigger Shift Is Toward Human-Centred Crash Engineering
Autoliv’s launch arrives as road safety remains a major public-health challenge. The World Health Organization’s July 2026 fact sheet estimates that approximately 1.16 million people die in road crashes each year, while another 20 million to 50 million suffer non-fatal injuries. Improving crash protection by even small amounts can therefore matter across millions of vehicles and years of exposure.
The new HBM platform also reflects a broader sharing of safety technology. The SAFER model originated through work involving Chalmers University of Technology, Autoliv and Volvo Cars and is now being developed and brought to market with the Fraunhofer-Chalmers Centre. Toyota, meanwhile, made its own THUMS technology freely available years ago. Those parallel efforts suggest that virtual-human modeling is evolving from a specialized capability held inside individual companies toward a wider technical ecosystem. Dummies will continue hitting barriers, sensors will continue collecting data and real vehicles will still be destroyed. But increasingly, the most revealing crash may happen first inside a computer—where engineers can see not only what happened to the car, but what might have happened to the person inside.
