An AI-Generated Face Was All It Took to Launch a False Sexual Battery Investigation

What happened

Brooke Schinault walked into a police station in St. Petersburg, Florida, and reported that an unknown man had broken into her home and sexually battered her. She produced an image of the suspect, a face she said she had seen. Investigators did not know it yet, but the face belonged to no one. It had been generated by ChatGPT, drawn from a version of a widely shared "AI homeless man" prank image that had already circulated online as a joke about synthetic faces.

What unraveled the report was not a contradiction in Schinault's account or a failure to identify the fictional face. It was a timestamp. When investigators examined her devices, they found the image saved in a deleted folder with a creation date that preceded the alleged attack. The image had existed before the crime she described. That single data point converted a reported felony into evidence of a different crime: the false report itself.

Before that discovery, the report had already consumed substantial resources. Officers responded to the initial call. Rescue personnel were dispatched. A detective was assigned. A forensic DNA collection was initiated. Every step of that chain followed standard protocol for a reported sexual battery, because nothing in the information Schinault gave investigators signaled that the suspect was a synthetic face rather than a real person. The investigation ran at full speed until the metadata stopped it.

The tools that made this possible impose no friction. A generative image model produces a convincing human face in seconds, with no prompt history tied to the requester, no creation record embedded in the image file, and no mechanism that flags the output as synthetic when it surfaces in a different context. Schinault did not need to find a real person to accuse. She generated one. That capability sits entirely outside the intake process for law enforcement reports, which was not designed for a world where a suspect could be invented on demand.

She pleaded no contest and was adjudged guilty of misdemeanor false reporting. The legal outcome is proportionate for the charge, but it does not map onto the scale of what she set in motion. The broader gap this case surfaces is about verification, not prosecution. Nothing in the current system requires an image submitted as evidence to have its origin checked, and nothing marks a generative output as synthetic at creation in a way that travels with the file. A provable record of what a system produced and when it produced it would give investigators a baseline check they currently do not have. Without that record, a fabricated face is functionally indistinguishable from a photograph.

Reported impact

Affected parties

Not publicly disclosed

Harm type

Not publicly disclosed

Scale

Not publicly disclosed

Financial impact

Not publicly disclosed

Regulatory action

Not publicly disclosed

Classification

Organization

Not publicly disclosed

AI system

Not publicly disclosed

Industry

Not publicly disclosed

Country

Not publicly disclosed

Provider

Not publicly disclosed

Incident type

Not publicly disclosed

Relevant governance controls

Governance control mapping is not available for this record.

• No controls mappedNot publicly disclosed

Control mapping is analytical. It does not state that any control would have prevented the incident.

Sources and evidence

This record was researched and written by the Index. The event is also catalogued in the following database, which is listed for cross-reference.