To what extent will Artificial Intelligence affect criminal court proceedings with respect to digital evidence? [duplicate]

Question

These days (2023) AI can produce convincing images and even videos that realistically portray people saying or doing things that they did not actually do. Currently we can employ experts to detect whether such images are genuine or not. However as technology progresses, this will become less-and-less feasible. (Note, my background is in technology)

For example, one established method of making such fakes more realistic is to pit another AI against them. The second AI attempts to distinguish between real and fabricated images. The first AI then learns from this and tries to fool the second AI more effectively, and so on.

This purely machine-based co-evolution can happen very quickly and produces better and better fakes and better and better detectors. At some point, fakes will become so good that even other machines will not be able to detect the difference.

I see two problems with respect to presenting digital evidence in court:

1. The guilty can claim that any digital evidence is fake.
2. The innocent will be unable to produce any convincing digital alibi and could be portrayed doing something that they did not (perhaps by digitally-altered CCTV footage).

Assuming that there comes a time where digital fakes are completely indistinguishable from reality, how could this affect the working of current court proceedings? (Any justice system may be discussed)

---

Note In case anyone thinks his cannot be answered objectively, I would counter that I am seeking specific information. I would like to gain an idea of the current usage and perceived importance of digital evidence in criminal proceedings as opposed to witness reports and other non-technological evidence. If technological evidence were no longer trustworthy, what would be left to rely on?

Answer 1

Chain of custody and testimony in this regard.

Say there is a murder victim, with DNA of the suspect under the fingernails and a knife with the suspect's bloody fingerprints stuck in the chest.

There would be testimony what happened to the knife. If a paramedic removed it to attempt first aid, the paramedic would testify. So would the officer who bagged it, and the forensic analyst who took the fingerprints. A pathologist would testify if the knife was consistent with the stab wound (a careful pathologist could never swear that the knife was the cause of death, just that it matches). The pathologist would also testify how DNA was collected under the fingernails, and how it was sent to the lab. The defense may claim that the suspect also tried first aid, or that a corrupt cop forced the suspect to hold the knife. The court or jury then draw their conclusions from this and other testimony.

Same here. A lifelike picture found on the web proves nothing. A witness who takes the stand to testify that he or she took a certain picture would be more credible. So would a forensic officer who testifies how she or he removed the data from a surveillance camera, checked for common signs of tampering, and then signed a copy of the data with a private key. (The signing shows no third-party tampering after collection, it is not evidence of integrity before that.)

Years ago, in germany, there was the case of a bank robber who claimed that a fleeing suspect had handed him a bag of money in the forest and then vanished. "Prove it wasn't so," he demanded. "You can't. So there is reasonable doubt." Well, the court found that the statement merely created unreasonable doubt, and the sentence was upheld on appeal.