Industries / Transport and vehicles / Trucking and freight / Fleet management
Drivers and safety
Drivers and safety is choosing, training and watching over the people who drive a fleet's vehicles, and cutting the crashes they are in.
6
Milestones achieved, from 1995 to 2024.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1995
Employers test their drivers
The Omnibus Transportation Employee Testing Act of 1991 required alcohol testing, on top of existing drug-testing rules, for people in safety-sensitive transport jobs, including anyone who drives a truck or bus on a commercial driver's license (CDL). By the account of FMCSA, the US truck and bus safety regulator, mandatory alcohol testing of these drivers began in 1995. The employer runs the tests: a drug test before hiring, and tests at random, after a crash and when a supervisor has reason to suspect use. A 2009 study of US fatal crashes from 1982 to 2006, which FMCSA cites, linked the testing programs to a 23% lower risk that alcohol was involved in a fatal crash by a commercial driver. A driver who failed a test could still be hired by a fleet that did not know, so from 6 January 2020 employers must check a federal database, the Drug and Alcohol Clearinghouse, before hiring a CDL driver and once a year for each driver they employ.
LessonTesting at one employer did little against a driver who moved to the next, until the results were pooled in one register.
Credit
the US Congress; the Federal Highway Administration and its successor FMCSA; Joanne Brady, Guohua Li and colleagues for the 2009 study
2012
A standard for road risk
In October 2012 the International Organization for Standardization (ISO) published ISO 39001, which sets requirements for a road traffic safety management system. It asks any organization that uses the roads, such as a company with a fleet, to stop treating road safety as a set of separate measures and to manage it as a system: focus on the factors it can control or influence, measure results and keep improving.
Credit
International Organization for Standardization
2017
Coaching beats warning lights
Researchers at NIOSH, the US federal workplace safety institute, fitted a commercial camera-and-sensor system to 315 vehicles at two companies, one in local trucking and one in oil and gas support work, and followed them for about two years. Warning lights in the cab alone did not cut risky driving more than in a group that got no feedback. Lights plus a supervisor who went over video of the driver's own risky moments did, and the drop lasted through about three months after the coaching ended. The camera maker supplied the equipment and coded the videos; NIOSH did not name it.
LessonThe camera alone changed little; the talk with a supervisor made the difference.
Credit
Jennifer L. Bell and colleagues at NIOSH, in the Journal of Safety Research
2021
London rates lorry windows
Since 1 March 2021 a heavy goods vehicle over 12 metric tons needs a safety permit to drive in Greater London. Transport for London (TfL) rates each lorry from zero to five stars by how much the driver can see directly through the cab windows, and lorries below the minimum must carry extra equipment such as side cameras and audible warnings. In the first year TfL issued 191,769 permits, 112,259 of them to zero-star lorries. Fatal collisions involving a lorry in which vision was a factor fell to six in 2021, from nine in 2019, by TfL's count, which TfL says must be read in the light of the pandemic; four of the six involved zero-star lorries. TfL planned to raise the minimum from one star to three from October 2024.
LessonA rating that buyers can see changes what fleets buy: hauliers told TfL they were building the star ratings into future purchases.
Credit
Transport for London; the Mayor of London
2023
Breath tests for company cars
Japan requires a business that uses five or more ordinary vehicles, or one vehicle with 11 or more seats, to check its drivers for alcohol before and after they drive, even when the vehicles do not carry goods or people for pay. The business must appoint a safe-driving manager and register that person with the police. From 1 December 2023 the check must be made with a working alcohol detector. That part was due in October 2022 but was put off because a semiconductor shortage left too few detectors, the news site Traffic News reported.
LessonA rule that depends on a device has to wait for the device, and the chip shortage held this one back by 14 months.
Credit
the Government of Japan, under the Road Traffic Act
2024
Speed warnings in new vans
Under the EU's General Safety Regulation, every new car, van, truck and bus registered in the EU from 7 July 2024 must have intelligent speed assistance, which reads the speed limit from road signs or map data and warns the driver when the vehicle is going faster; new vehicle types needed it from 6 July 2022. The EU chose a version the driver can override, according to the European Transport Safety Council (ETSC), a research and advocacy group.
LessonThe US went the other way for heavy trucks: on 24 July 2025 its truck and vehicle safety regulators, FMCSA and NHTSA, withdrew a 2016 proposal to require speed limiters, citing "continued data gaps" about costs and benefits.
Credit
the European Parliament and Council (Regulation 2019/2144); the European Commission
Watch for
Dash cameras cut crashes
Samsara says fleets of 175 or more vehicles that used its full set of AI safety tools, dash cameras facing both the road and the driver, alerts in the cab and coaching, saw their crash rates fall about 75% over 30 months, and that its customers typically see crashes fall 35% to 40% over time. The figures come from its own customers' data, more than 2,600 fleets by FreightWaves' account, with no comparison against fleets that did not buy the system.
Achieved whenan independent researcher compares the crash records of fleets that adopted such cameras with matched fleets that did not, and publishes the method.
Claimed by
Samsara, October 2025
Open
01
Count work-related road deaths
An open yearly table of road deaths that involved someone driving or riding for work, country by country, built from national crash data with one published definition, together with the code that rebuilds it.
The briefCredit · unclaimed02
Check camera crash claims
An independent before-and-after study of US trucking companies that have said in public that they adopted AI dash cameras, compared with similar companies that have not, using public federal crash records.
The briefCredit · unclaimed
Open 01 · the brief
Count work-related road deaths
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open yearly table of road deaths that involved someone driving or riding for work, country by country, built from national crash data with one published definition, together with the code that rebuilds it.
Why it is still open
In December 2025 ETSC reported that Europe has no common definition of a work-related road collision: 20 of the 30 countries it covered have one and 10 do not, and only 13 record the purpose of the journey in police reports. Reported shares of road deaths range from 42% in France to 2% in one country.
Built with what exists
national crash data from the countries that record journey purpose, as listed by ETSC; the US FMCSA Crash File, which records crashes of commercial vehicles; Python and pandas; an AI model such as Claude to read each country's statistics manual and log its definition, checked by a person.
First step
Pick two countries that publish crash records with journey purpose and count one year of work-related deaths under each country's own definition and under one shared definition.
Done looks like
15 countries and five years each counted under one definition, the method and code published, and for each country the gap between its own reported figure and the common-definition figure.
Open 02 · the brief
Check camera crash claims
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An independent before-and-after study of US trucking companies that have said in public that they adopted AI dash cameras, compared with similar companies that have not, using public federal crash records.
Why it is still open
The published crash-reduction figures come from vendors' own customer data, such as Samsara's 2025 report. The best-known side-by-side test, by the Virginia Tech Transportation Institute (VTTI) in 2023, was paid for by one vendor, Motive, measured alerts on a test track rather than crashes, and was disputed by a rival, Lytx. The NIOSH study measured risky driving, not crashes. The search for this file found no independent study of crash outcomes.
Built with what exists
FMCSA's public Crash File, which lists crashes by each carrier's USDOT number; FMCSA's carrier census files, which give each company's number of vehicles and drivers; carriers named in vendors' public case studies, gathered by a web agent (see the Note "Deleting 80% to start over"); Python, pandas and a standard matching method such as propensity scores.
First step
Download the Crash File, pick ten carriers named in vendors' case studies and chart their crash counts by year.
Done looks like
100 named adopting carriers matched to 300 others by size, type of work and state, with crashes per truck for two years before and two years after adoption, an analysis plan published before the results, and the list and code released.