When Algorithms Become Evidence: A Warning on AI Deepfakes and Due Process

By Michael Phillips | NYBayNews

A troubling case tied to Pennsylvania is sounding a national alarm about how quickly artificial intelligence is outpacing the justice system — and how ordinary Americans can lose their freedom when courts fail to slow down and verify digital evidence.

According to a January 9, 2026 Action News Investigation originally reported by WPVI-TV in Philadelphia and syndicated nationally, Melissa Sims, a Delaware County native, was wrongfully jailed after allegedly AI-generated fake text messages were accepted as real by authorities — without verification.

The result: two days in jail, eight months of legal limbo, and a stark lesson in what happens when technology outruns due process.

A Case That Should Never Have Happened

Sims’ ordeal began after a domestic dispute in Florida in late 2024. Although she called police alleging her boyfriend had vandalized their home and injured himself, she was arrested for domestic battery. As part of her bond, a judge imposed a strict no-contact order.

Months later, Sims was arrested again — this time for allegedly violating that order.

The evidence? Text messages her ex-boyfriend claimed came from her.

Sims says they were fabricated using AI tools to impersonate her writing style and phone number. Despite growing awareness of AI-generated deepfakes, no forensic analysis was conducted. No metadata review. No authentication. No expert verification.

“No one verified the evidence,” Sims told investigators.

She was jailed in general population for two days before her attorney began unraveling the claim. After eight months, prosecutors dropped the bond violation. In December 2025, Sims was acquitted of the original battery charge as well.

The damage, however, had already been done.

The Broader Threat: Unverified AI in Low-Threshold Cases

What makes this case especially concerning is where it occurred procedurally.

Bond violations, protective orders, and domestic cases often rely on fast decisions, limited hearings, and informal digital evidence. That makes them uniquely vulnerable to abuse — especially when AI-generated texts, images, or audio can be produced in minutes.

Experts interviewed in the investigation warned that detection tools lag far behind generation tools. One AI forensics professor demonstrated that the same fake image could be labeled anywhere from 1% to 62% “likely AI-generated,” depending on the software used.

The implication is clear: courts are trusting evidence that even experts can’t reliably authenticate.

Pennsylvania Acted — But the System Hasn’t Caught Up

In July 2025, Pennsylvania took a step in the right direction when Governor Josh Shapiro signed a digital forgery law making it a felony to create harmful AI deepfakes intended to injure, exploit, or defraud others.

That law recognizes the threat.

But Sims’ case exposes a deeper problem: criminalizing misuse after the fact doesn’t protect people when judges, prosecutors, and police accept AI-generated content at face value in the first place.

From a center-right perspective, this raises serious concerns:

• Limited government should not mean careless government
• Due process must not be optional because a case is “low-level”
• Technology should not become a shortcut around constitutional protections

If anything, the lower the stakes procedurally, the higher the burden should be on verification — because defendants have fewer safeguards.

A Case for Verification Standards

This is not an argument against technology. It is an argument against blind trust in it.

A responsible, liberty-minded approach would include:

• Mandatory authentication standards for digital evidence
• Clear rules requiring forensic verification when AI manipulation is alleged
• Penalties for knowingly submitting fabricated digital evidence
• Training for judges and law enforcement on AI risks

As one expert put it in the investigation, the justice system must move from “trust until disproven” to “distrust until verified.”

That principle aligns with core conservative values: skepticism of unchecked systems, protection of individual liberty, and insistence on accountability.

If It Can Happen to Her, It Can Happen to Anyone

Sims is now advocating for stronger laws and evidence standards, warning that her case is not an anomaly — it’s a preview.

In an era where a convincing fake text can be generated faster than a warrant can be reviewed, the question is no longer whether AI will reach the courtroom.

It already has.

The real question is whether the justice system will adapt before more innocent people lose their freedom to an algorithm — and a lack of verification.