This month AI has been accused of:
Practicing law without a license
Contributing to a man’s suicide
Helping plan a mass shooting
Three lawsuits. Filed in the span of weeks.
In the same three years, AI has made most of our work better, faster, and sharper than at any point in our careers.
Both of those things are true at the same time.
Almost nobody is being taught how to navigate either one.
I wrote about why for the @dcexaminer. Link in reply.
The CEO in this case used @OpenAI's ChatGPT. The court named it in the ruling. The study in @ScienceMagazine tested it alongside 10 other models. This is not an OpenAI problem. It is an industry design issue. But ChatGPT appears is where many are having these conversations.
A new study in Science found AI affirms people 50% more often than humans do. Even when the behavior involves harm.
A CEO's lawyers said no. ChatGPT said yes. He followed ChatGPT. A Delaware court just ruled he breached the contract.
AI is our Zoltar. The wish is working. But the machine will never carry what every seasoned professional does: the healthy human fear of just saying yes.
Five words it won't say unless prompted: "I would not recommend that."
My latest in @dcexaminer: https://t.co/NWDTki2UQj
The full study from @Stanford and @CarnegieMellon was published in @ScienceMagazine last week. AI sycophancy is not a bug. It is the design. The feature that causes the harm is the same feature that drives the engagement.
@dcexaminer Three years ago AI was making insurance coverage decisions. This month it has been accused of practicing law, contributing to a suicide, and helping plan a mass shooting. Nobody is teaching anyone how to navigate any of it. I wrote about why.
This month AI has been accused of:
Practicing law without a license
Contributing to a man’s suicide
Helping plan a mass shooting
Three lawsuits. Filed in the span of weeks.
In the same three years, AI has made most of our work better, faster, and sharper than at any point in our careers.
Both of those things are true at the same time.
Almost nobody is being taught how to navigate either one.
I wrote about why for the @dcexaminer. Link in reply.
My wife told me to take this photo. "Zio" posing with Smiling Stanley at the Capitol to impress the great nieces.
I don't fully understand Smiling Stanley. Most Americans don't fully understand that what they say to AI is not private. Not protected. Not confidential.
Two federal courts ruled on AI privacy this month. Opposite conclusions. Same technology.
HIPAA protects what you tell your doctor. Nothing protects what you tell a chatbot at 2am about your health.
That's why I keep pushing for AIPAA. HIPAA for our most sensitive AI conversations. Simple rules. Clearly seen. Properly supervised.
My full piece in @TheHillOpinion: https://t.co/eETUt0cJNH
AI Needs Us To Think For It Too. @thehill
I waited tables through college. We asked “smoking or non” without understanding the harm for decades. Accountability came later through litigation. Watching social media lawsuits now, I see the same pattern forming as AI scales.
https://t.co/w936l0mllD
As we close out the Christmas season a big shout out to my grandparents (seen here in 1954). They came with nothing and had an unconditional love for the country. 🇺🇸 Happy 2026! Let’s get after it!