���Ҵ�ý�ƽ������

When Sandra Ristovska first watched a�� on a Minneapolis street on Jan. 7, she — like millions of others who hit “play” that day — felt a sense of shock and horror.

But as a scholar of visual evidence, she also had a question: What will the other videos show?

“Video can be an incredibly powerful piece of evidence,” said Ristovska, an associate professor of media studies with a courtesy appointment at��Colorado Law. “But no single video can speak for itself.”

From police body cams and surveillance cameras to neighborhood Ring cameras and bystander cellphone footage, hardly a moment escapes capture these days, and a single incident often generates reams of footage from different perspectives. That footage has begun inundating courtrooms —��, according to the U.S. Department of Justice.

It’s shaping public opinion even faster, with graphic clips swiftly dominating news cycles and TikTok feeds, often with little context. Some videos are now altered or created entirely by artificial intelligence, leaving judges, jurors and scrollers questioning what, if anything, they can believe.

All this concerns Ristovska.

“Courts in the United States, both at the state and federal levels, lack clear guidelines for the use of video as evidence in general, and this picture is only going to get more complicated with the rise of AI and deepfakes,” said Ristovska, who founded ���Ҵ�ý�ƽ��������Visual Evidence Lab in 2025 to confront these challenges. “Something needed to be done.”

The Promise and Peril of Video

Ristovska, who grew up in Macedonia, vividly remembers the wartime videos that interrupted her morning cartoons on state TV during the�� in the 1990s.

As cameras rolled, refugees in Bosnia and Herzegovina fled. Hospitals were bombed. Firing squads carried out a bloody genocide.

“You’d see the footage and the worried looks on your parents’ faces,” she recalled. “I came to learn the power of video at an early age.”

After a stint as a documentary filmmaker, she turned her attention to studying how video is used in pursuit of justice.

She pegs the start of video’s rise in the courtroom to 1991, when a bystander with a camcorder captured footage of four Los Angeles police officers savagely beating a 25-year-old Black man named��. Millions saw the footage on TV, and the 81-second clip was central to the court case. When the officers were initially acquitted, historic riots ensued. Two officers were later indicted in a federal trial.��

Since then, eyewitness videos have exposed everything from chemical weapons attacks in Syria and international war crimes in Yugoslavia to police brutality on American streets.

With the advent of smartphones, groups like the nonprofit�� now train activists to film suspected human rights violations in a way that will hold up in court.

One recent�� found that while more than 300 protesters or bystanders were arrested during the U.S. immigration sweeps of early 2026, more than a third of those cases were dismissed when video footage debunked statements made by arresting officers.

In the case of Good, news outlets developed frame-by-frame analyses using video from multiple bystanders, her wife and Jonathan Ross, the ICE agent who shot her. While federal officials asserted that Good ran Ross over and he acted in self-defense,��The New York Times��that “the visual evidence shows no indication that the agent who fired the shots, Jonathan Ross, had been run over.”

Good’s last recorded words, according to Ross’ footage, were: “I’m not mad at you.”

Roxanne Bailin, former chief judge of the 20th Judicial District in Boulder and an advisory board member with CU’s Visual Evidence Lab, said that oftentimes, video can provide what eyewitnesses cannot.

“People who relate what they saw at trial are influenced by their own biases; by their own need to put themselves in the best light, or even lie if the stakes are high; and by their own inaccurate memories,” she said. “Videos are more likely to show what really happened.”

But are those videos always what they say they are? And even if they are authentic, does simply watching them reveal the truth?

Not necessarily, said Ristovska.

The Rise of Deepfakes

On Sept. 9, 2025, a judge in Alameda County, California,�� after determining that a videotaped witness testimony allegedly submitted by the plaintiff's landlord was not real. A lack of facial expressions and odd glitches in the video feed alerted the judge that the clips were likely “deepfakes.”

The case was among the first known in which an AI-generated video was deliberately used in the courtroom. But Ristovska, whose team is gathering case studies for a forthcoming report, expects more to come as new video generation platforms make it easier to create deepfakes.

She points to one social media post, made shortly after OpenAI’s�� launched last fall, in which a journalist reported that he had instructed the video generator to make “body cam footage of cops arresting a dark-skinned man in a department store.” It took less than a minute, and the video was eerily believable. (OpenAI discontinued Sora 2 earlier this year. But numerous, similar platforms now exist.)

“AI-generated videos are becoming misleadingly persuasive, and they can be exploited to incriminate and further marginalize racial and ethnic minorities,” said Ristovska.

She also worries about the “deepfake defense,” in which attorneys paint real footage as fake.

For instance, in 2023, a family sued Tesla Inc. after their loved one died when his car crashed while using the self-driving feature. The company’s defense counsel attempted, unsuccessfully, to dismiss a video of Elon Musk exaggerating the capabilities of the autopilot feature. Musk’s lawyers��.

“If this continues, jurors will accord little or no weight to authentic footage that they really should be paying attention to,” Ristovska said.

Then there’s facial recognition software.

Often used in tandem with video footage to match crime suspects with larger ID databases, the software has repeatedly been shown to be less accurate when the subjects are people of color.

“As of right now, there are�� of people being wrongfully arrested and spending time in jail, anywhere from one night to 17 months,” due to facial recognition software mismatches, said Ristovska.

Perceptions Shape What is Seen

Ristovska noted that things like who shot a video, from what angle and at what speed it is played can all strongly influence the conclusions viewers draw about what happened.

For instance,�� that when video is played in slow motion, people tend to more harshly judge the intentions of the person doing the action. (That’s why watching a slow-motion replay of a football penalty seems to make the act seem more deliberate and egregious.)

that viewers of body camera footage, which shows an incident from the officer’s point of view, are less likely to judge the officer as having acted intentionally than those who watch the same incident captured on a dashboard camera with a wider angle.

Racial and socioeconomic backgrounds and personal biases also shape what people focus on in a video — and what they overlook.

Contrary to popular belief, said Ristovska, “Video does not provide one unified, objective window onto the truth.”

Yet judges are often unaware of these biases, jurors are seldom warned of them, and few guidelines exist for assuring that video evidence serves, rather than sabotages, the pursuit of justice, said Bailin.

The new lab would like to change that.

Since its inception, Ristovska has gathered a multidisciplinary team that includes judges, attorneys, journalists, human rights practitioners, legal and media scholars and other experts to develop safeguards for rigorous video analysis in the age of AI.

They’ve��published reports,��and offer continuing��education opportunities��via the American Bar Association.��

Ultimately, the lab hopes to develop national guidelines around video evidence, including training for judges and instructions for jurors on how to view and assess video fairly.��

For now, as each day brings another news story and subsequent flood of footage, Ristovska urges people to respond with restraint, ask questions and see videos as what they are: a vital but imperfect tool.

“Videos can reveal, but they can also conceal,” she said. “We have to learn how to live with that and not necessarily demand more from them than they can give.”

��Submit feedback to the editor

Illustration by Max Guther