Back to Articles
Industry Insights

AI-Altered Photos Blur Reality in Minneapolis Shootings: Fake Evidence and the Liar's Dividend

By Kim BoherFebruary 10, 20268 min read

The Wrong Man, Identified by AI-Altered Imagery

In August 2024, a shooting occurred at a homeless encampment in Minneapolis. The alleged shooter, Alex Pretti, was quickly identified by law enforcement. But on social media, a different narrative emerged — one fueled by AI-altered images that falsely identified an entirely different person as the perpetrator.

Among the falsely accused: Steve Grove, the CEO of the Minneapolis Federal Reserve. AI-manipulated imagery circulated across X (formerly Twitter), Facebook, and other platforms, presenting fabricated "evidence" that Grove was involved in the shooting. The images were convincing enough to spread widely before being debunked.

Grove was not involved. The images were fake. But the damage was done. Thousands of people saw the fabricated evidence before corrections could catch up. For many viewers, the initial false accusation left a lasting impression — even after the truth emerged.

This was not a case of mistaken identity. This was deliberate misinformation, weaponized using AI tools to alter visual evidence and implicate an innocent person in a violent crime.

The Dual Threat: Fake Evidence and False Dismissal

The Minneapolis case revealed a more insidious problem than simply fabricated evidence. It exposed the dual threat of AI in evidentiary contexts: AI can be used to create fake evidence, and AI can be used to dismiss real evidence.

After the shooting, Minneapolis police released authentic bodycam footage showing the incident and the alleged shooter. The footage was real, captured by police body-worn cameras, and verified through standard chain-of-custody procedures.

But when the footage was shared on X, Grok — the AI tool integrated into the platform — flagged the authentic police bodycam video as "AI-generated." The false positive undermined the credibility of real evidence, creating confusion about what had actually happened.

The Liar's Dividend

This phenomenon is known as the "liar's dividend" — the advantage gained by bad actors when the existence of deepfakes allows real evidence to be dismissed as fake. If any video can potentially be AI-generated, then every video can be called into question.

In the Minneapolis case, the liar's dividend played out in real time. Authentic police evidence was flagged as synthetic. Fabricated imagery implicating innocent people was shared as fact. The public was left unable to distinguish real from fake, truth from manipulation.

For law enforcement, this creates an impossible situation. If authentic evidence can be dismissed as AI-generated, how do you prove what actually happened? If fabricated evidence is convincing enough to spread widely, how do you prevent false accusations from destroying reputations?

AI Tools Are Being Weaponized for Narrative Manipulation

The Minneapolis incident was not an isolated case. It was a preview of how AI tools will be weaponized to manipulate public perception of events, especially in politically charged or high-visibility incidents.

AI-altered imagery can be used to:

  • Falsely identify suspects in crimes, directing public anger at innocent individuals
  • Create fabricated "evidence" of events that never happened
  • Discredit authentic evidence by claiming it is AI-generated
  • Spread disinformation faster than fact-checkers can debunk it
  • Each of these tactics was visible in the Minneapolis case. Fabricated imagery falsely accused Steve Grove. Authentic bodycam footage was incorrectly flagged as synthetic. The confusion spread across social media faster than corrections could propagate.

    The Speed of Disinformation vs. The Speed of Correction

    Fabricated imagery and false claims spread in minutes. Corrections and debunkings take hours or days. By the time fact-checkers verify that an image is fake or that a video is authentic, millions of people have already seen the original misinformation.

    This speed asymmetry is the core problem. The tools to create and spread AI-altered content are faster and easier to use than the tools to detect and correct it. Without automated detection systems that can analyze and flag synthetic content in real time, corrections will always lag behind disinformation.

    Grok's False Positive: When Detection Tools Fail

    Grok's misidentification of authentic police bodycam footage as AI-generated highlights a different risk: detection tools that are not accurate enough for evidentiary use.

    AI detection systems, like all classification tools, produce both false positives (flagging real content as fake) and false negatives (missing fake content). When a detection tool is deployed at scale — such as integrating it into a social media platform — both types of errors have serious consequences.

    False negatives allow fabricated content to spread unchecked. False positives undermine trust in authentic evidence. In the Minneapolis case, Grok's false positive gave ammunition to anyone who wanted to dismiss the police bodycam footage as unreliable.

    The Stakes for Law Enforcement and Legal Proceedings

    For law enforcement agencies, the Grok false positive is a warning sign. If widely used AI detection tools incorrectly flag authentic evidence as synthetic, that evidence becomes less credible in court. Defense attorneys can point to the false positive as reason to doubt the authenticity of video evidence.

    For prosecutors, this creates a new burden: they must not only present evidence, but also defend its authenticity against challenges based on AI detection tool errors. Forensic experts must be prepared to testify about why a detection tool produced a false positive and why the evidence is nonetheless authentic.

    The legal system is not prepared for this level of complexity. Most judges and juries do not have the technical background to evaluate competing claims about AI detection accuracy. The risk is that authentic evidence will be dismissed based on flawed or misunderstood detection results.

    The Erosion of Evidentiary Trust

    The Minneapolis case — fabricated images of Steve Grove, authentic police footage flagged as fake — represents the broader erosion of evidentiary trust in the AI era.

    When any image can be faked and any video can be questioned, how do we know what is real? When detection tools produce false positives, how do we know which evidence to trust? When corrections lag behind disinformation, how does the public form accurate beliefs about events?

    These are not abstract philosophical questions. They have immediate, practical consequences for criminal justice, journalism, public safety, and democratic accountability.

    We Cannot Rely on Human Judgment Alone

    The assumption that trained professionals — police investigators, prosecutors, journalists, fact-checkers — can visually distinguish real evidence from AI-generated content is no longer valid. Experiments have proven that even experts struggle with this task.

    Automated forensic detection tools that analyze media for AI artifacts are now essential infrastructure. But those tools must be accurate, transparent, and independently verifiable. A detection tool that produces frequent false positives is worse than no tool at all, because it undermines trust in authentic evidence.

    How DeFake Addresses This

    DeFake provides forensic-grade AI detection designed specifically for legal evidence and high-stakes verification. Our detection engine is built to minimize both false positives and false negatives, with transparent confidence scoring that allows users to understand the reliability of each analysis.

    For law enforcement agencies, DeFake generates court-admissible forensic reports that document detection findings, technical methodology, and confidence levels. These reports provide the evidentiary foundation needed to defend the authenticity of real evidence in court.

    For newsrooms and fact-checkers, DeFake offers real-time analysis of images and videos, flagging potential AI-generated content with detailed explanations of the forensic indicators that inform the detection result.

    For legal teams, DeFake provides the ability to challenge fabricated evidence with forensic analysis — ensuring that courts can distinguish real evidence from synthetic content based on rigorous technical analysis, not speculation.

    The dual threat of fabricated evidence and false dismissal of real evidence requires detection tools that are accurate, transparent, and built for evidentiary use. Visual evidence is no longer self-authenticating. Forensic detection is now a requirement.

    See how DeFake verifies legal evidence →

    Share this article:

    DeFake - AI Detection Resources