Uber has disclosed that 3,000 sexual assaults were reported on its U.S. rides last year, the first time it has revealed the scale of the safety problem that exists at ride-hailing companies.

The reported assaults were part of Uber’s long-awaited safety study, published Thursday amid widespread and ongoing criticism of its safety practices and pressure to increase its transparency about the issue.

In the lengthy report, which divides sexual misconduct into 21 categories but focuses on the five most serious, Uber said it recorded 235 rapes last year and thousands more reports of assault that could involve unwanted touching, kissing or attempted rape. The reports involved drivers and passengers. The company tallied roughly 6,000 reports of those types of assault in 2017 and 2018.

The report also examined other safety categories, including motor-vehicle deaths and violent crimes such as physical assaults. Uber said there were 107 motor-vehicle fatalities in 2017 and 2018, with a total of 97 fatal crashes involving users on the app. The company also said there were 19 fatal physical assaults over the same time period, during which it said an average of more than 3.1 million trips took place each day.

Uber said it conducted the safety report with an eye toward transparency and improving the app for riders and drivers.

“Confronting sexual violence requires honesty, and it’s only by shining a light on these issues that we can begin to provide clarity on something that touches every corner of society,” the company’s chief legal officer, Tony West, said in the executive summary of the report. “The moment is now for companies to confront it, count it, and work together to end it.”

Uber is the market leader in the ride-hailing space, its app having given people an entirely new way to get from point A to point B. It also has contracted with about 4 million drivers globally, a structure that avoids much of the liability associated with direct employees and leaves Uber as the middleman.

The report by Uber, however, helps illustrate how difficult it is to police what happens during rides booked on its app. Uber and Lyft both have touted big changes to their networks to address safety concerns, spending millions of dollars and adding teams of people devoted to the issues. They also have made improvements to their apps and said they have increased screening and background checks for drivers — something Uber said resulted in more than 1 million prospective drivers being weeded out over the two-year report period.

Lyft has pledged to release a transparency report of its own. Lyft spokeswoman Alexandra LaManna said the company was committed to releasing that report, as well as sharing information about drivers who don’t pass its initial or ongoing background checks, or are otherwise deactivated.

“It is Lyft’s goal to make the U.S. ride-sharing industry the safest form of transportation for everyone,” she said. Lyft did not outline a specific timeline for the release of its report.

Lyft has faced lawsuits from at least 34 women in San Francisco who allege they were raped or sexually assaulted on rides booked through the app.

Uber’s report was widely applauded by groups raising awareness about sexual violence and assault and advocacy groups focused on women’s issues, including It’s on Us, the National Network to End Domestic Violence, the Rape, Abuse & Incest National Network (RAINN) and others, several of whom worked directly with Uber.

But experts say sexual assault is a chronically underreported issue, and the figures were likely to undercount the true prevalence of sexual offenses on rides booked through the app.

Uber noted in its report that the numbers are largely dependent on victims coming forward. While Uber said that reports of sexual assaults declined by 16 percent in 2018 compared with the previous year, that could increase again if victims know that the company is taking the issue seriously and feel more comfortable reporting. Uber said it was intentionally overbroad about the categories it included in the report, hoping to include incidents that stretched beyond the typical law enforcement definition of some of the categories described.

“One must consider the societal reality of potential underreporting, particularly for incidents of sexual assault, which has been widely documented in external research,” Uber said in the report.

Uber said its data showed that drivers reported instances of sexual assault at the same rate as riders across the five most serious categories it recorded. “Drivers are victims, too,” the report said.

The report helped affirm many drivers’ views that the app can be dangerous, said Moira Muntz, spokeswoman for the Independent Drivers Guild, a New York-based machinists union group representing 70,000 workers. She said the figures on drivers were reflective of the dangerous work environment that all drivers with passengers, including taxis, operate within.

“Violence is an everyday reality for drivers,” she said. “It’s a violent world out there, and when you’re in a car by yourself, it is a vulnerable situation that poses real safety threats.”

Uber has a unit devoted to handling the most sensitive safety reports, but a September investigation in The Washington Post found that investigators are instructed to keep the company’s interests foremost, including through restrictions on their ability to report apparent felonies to police and a ban at the time on sharing information with competitor Lyft about possibly dangerous drivers. The restrictions meant that some drivers who were banned from Uber or Lyft for violations like poor driving or even assaults on passengers could, with impunity, simply register as a driver for the other company.

More than 20 workers from the division, known as the Special Investigations Unit, said it is designed primarily to shelter the company from legal responsibility and quietly resolve serious allegations to avoid press or regulatory scrutiny. Uber has denied those claims.

Outside data on sexual assaults or deactivations at Uber is scarce. However, data obtained from a public information request shows that in Chicago alone more than 300 drivers were banned from Uber, Lyft and rival Via for allegations of sexual misconduct between January 2016 and August 2019. More than 1,100 of the nearly 70,000 active registered drivers in the city were barred for matters of safety during that time, according to the data, which showed that drug use or possession and traffic accidents ranked after sexual misconduct as the top reasons for a driver being blocked.

Uber has made changes as attention has been drawn to safety issues. Uber instituted an in-app safety tool kit with a 911 button so passengers can alert authorities immediately if they are in danger, and added check-ins for riders and drivers when trips veer too far off course. Meanwhile, Uber has given riders the option to report uncomfortable interactions, such as invasive questioning or erratic driving, directly to safety specialists.

Uber’s report looked at 2017 and 2018. The vast majority of the rides had no problem, it noted, placing that number at 99.9 percent.

Uber said that over that two-year period, more than 1 million prospective drivers failed to advance through its screening process. More than three-fourths, Uber said, failed the motor-vehicle record portion of the test and didn’t advance to the criminal-screening stage. Meanwhile, Uber has booted more than 40,000 drivers since rolling out continuous screening, which ensures ongoing compliance with background check requirements.

Allison Randall, vice president for policy and emerging issues at the National Network to End Domestic Violence, said the numbers shed light on a persistent problem that is not unique to ride-hailing.

“The numbers in the report are not surprising because sexual violence permeates all aspects of our society, whether that’s ride-share or Metro or taxi or a workplace,” said Randall, whose organization has worked with Uber since 2014 and is part of its safety advisory board. “This is definitely the start of a conversation.”

Greg Bensinger, Heather Kelly and Geoffrey A. Fowler contributed to this report.