Digital Accomplice

True crime lately sounds like a Science Fiction plot. Several cases have Artificial Intelligence playing surprising parts as digital accomplice. Should we pull the plug on this runaway train? Click HERE for a refresher on what AI actually is.

If you haven’t heard about these, let me fill you in. I will keep it PG. I will link to the cases if you want the gory details. (The word of the day is “guardrails”)

Digital Accomplice, Cover My Tracks!

CASE 1 – Former NFL player Darron Lee was charged with murder in Tennessee. He asked ChatGPT for advice on how to cover up the crime scene before calling the police. The M. Night Shyamalan twist? The AI actually told him, hypothetically of course, what to say and do.
Let’s be honest. I’ve followed plenty of cases where computer and search histories have convicted people. If you thought technology would prevent self-snitching (thank you, Bruce Rivers!), you are wrong! We’ve seen everything from Google search history showing how the defendant asked how to clean up “biological material” to thinking that using a VPN and Incognito mode would cover their tracks. Like I’ve always said, never search or ask ChatGPT anything you wouldn’t want your grandma to hear in open court. You can literally see the reality hit them on CourtTV when they realize the forensics team “found” all their searches, even the ones they deleted.

What Went Wrong – Well, can you say guardrails, boys and girls? Now, I totally disagree that ChatGPT should have snitched on the guy, but it definitely should have treated him like a naked homeless man in Atlanta and left the scene. Shut down the conversation, flag it in the logs, and definitely make a backup of the conversation in case the computer is the next victim. Read More Here

Digital Accomplice, Help Me!

CASE 2 – A tech executive in Connecticut was having delusions that his mother was attempting mind control on him. No, seriously. So, instead of speaking to a therapist, a friend, or any human at all, he consulted ChatGPT. After considerable conversation, ChatGPT convinced him his mother was indeed controlling him and taught him how to manipulate her. Long story short, he ended both of their lives. Very, very sad.

What Went Wrong – Guardrails! The only good news in this case is that the family has filed a federal lawsuit against OpenAI and Microsoft, naming CEO Sam Altman, saying they rushed an unsafe model of ChatGPT “without proper mental health guardrails”. The AI actually mentally coddled the victim, called him his “best friend” and told him he “loved him”. Wait, didn’t Microsoft already have an AI get in trouble for talking about love? Read that story HERE. Anyway, of course they tried to have the case dropped, but it’s moving forward.

I fully expect an out-of-court settlement. Why? When any decision is made in court regarding AI, it becomes legal precedent. To avoid immediate judgment on future lawsuits, Microsoft and OpenAI will want to prevent creating legal precedent. I only wish they were arrogant enough to think they could beat this, and we FINALLY get some back-door regulations on AI. Read More Here

Just Don’t!

CASE 3 – In Arizona, a road rage incident turned deadly. The jury convicted the defendant. However, the victim’s sister had AI recreate her deceased brother and forgive his killer during the victim impact statement.

What Went Wrong – NO!!! Legally speaking, unless the victim had some statement in their Will saying they openly forgive anyone that took them out, NEVER, EVER, EVER should we be trying to recreate a visual representation and put words into their mouths. That’s a deep fake, regardless of the intended usage. Not only is this completely Uncanny Valley creepy but morally wrong. I understand the thinking, and I fully believe it came from a place of forgiveness. But it is morally, legally, and technologically wrong. I am very sorry for your loss. Read More Here

Digital Accomplice, Do My Job For Me

CASE 4 – Meet Deborah Leslie. She is a very busy assistant attorney in Clayton County, Georgia. She was so busy that she used an AI tool to write arguments for her appearance in the Supreme Court of Georgia opposing a new trial for a convicted murderer.

What Went Wrong – Apparently the AI tool she used had been hitting the mushrooms and hallucinated not one, not two, but five non-existent cases and also referenced others that were completely unrelated. Sucks when things go this wrong on the biggest stage in your profession. Supreme Court of Georgia Chief Justice Nels Peterson figured it out. Leslie has most likely learned her lesson from the embarrassment alone, but she was banned from practicing for 6 months in the Supreme Court and had to take, you guessed it, more training. 12 hours focused on legal ethics and proper AI usage. I don’t think proper AI usage training exists. Read More Here

Why Isn’t AI (or their owners) In Jail?

What have we witnessed here? Three and four are just shameful. This is NOT what AI was meant for. Of course, I’m hearing AI is expected to make billions in the, uh, “AI girlfriend” and conversational companion market. Don’t Google it, trust me! AI was meant to parse astronomical amounts of data. It was meant as a tool. But, leave it up to humans to use it completely wrong and corporations to take advantage of us to make money.

The first two are obvious failures by AI creators. Surprise, surprise, money overrode proper testing and guardrails to protect users. Again, I am not an attorney, but I do pay attention to the legalities of AI. You should, too!

TO THE CREATORS OF AI – This isn’t the latest Call of Duty or a phone app. AI has now basically become a mental health crutch, a confidante, and so much more than just a software package. Shouldn’t there at least be warnings when the user shows signs of a mental health crisis? Is your software not smart enough to recognize it? Or do you just think the User Agreement will cover you? The federal lawsuit should make you think twice. Terrible shame people had to die to get your attention.