This week showed what happens when AI meets real consequences. An expert witness's ChatGPT prompts ended up in a court record, and a supermarket's face scanner wrongly flagged a customer. A new survey found most Americans now worry more than they hope about AI, even as business spending keeps rising and a Chinese robot maker had one of the year's biggest stock debuts.
An expert witness used ChatGPT to write a report defending 3M
An expert hired by 3M used ChatGPT to write large parts of his report in a lawsuit over a deadly 2020 explosion at a Houston manufacturing plant, 404 Media reported on August 17. His prompts asked the chatbot to show the company was not at fault. Court transcripts and discovery records show those prompts can be pulled into litigation, the outlet reports.
Our take: Your prompts are business records. Anything your team types into an AI tool could one day be read by a judge, a regulator or the other side, so write prompts you would be comfortable seeing in public.
A supermarket's face scanner wrongly accuses a shopper
A Sainsbury's store in the UK paused its AI face-scanning system after it wrongly flagged a paying customer as a shoplifter, the Guardian reported on August 17. The system was meant to spot known shoplifters as they walked in. Instead it pointed staff at an innocent person who had paid.
Our take: An AI mistake made in front of a customer becomes your mistake. If you use AI to make calls about people, from fraud flags to hiring screens, a trained person should check before anyone acts on it.
Most Americans now worry more than they hope about AI
A new Pew Research Center survey found that 52% of US adults are more concerned than excited about AI, and for the first time that includes adults under 30, The Register reported on August 19. Seventy-one percent expect AI to reduce US jobs over the next two decades, up from 64% two years earlier. Only 9% are more excited than concerned.
Our take: Your customers and staff are uneasy, even if they use AI daily. Be open about where you use it, and tell employees plainly how it will and won't change their jobs.
OpenAI gains on Anthropic among business buyers
Anthropic led OpenAI in share of paying business customers in July, at nearly 44% against nearly 40%, but OpenAI is growing faster so far this quarter, according to Ramp spending data reported by TechCrunch on August 20. The share of Ramp's customers paying for any AI rose to nearly 56% in July. Ramp's data covers more than 70,000 US businesses.
Our take: More than half of businesses in that sample now pay for AI, so it is becoming a normal line item. With two strong vendors close together, you have room to negotiate and no reason to lock into one.
Unitree's stock jumps more than 460% on debut
Chinese humanoid robot maker Unitree closed its first day on Shanghai's STAR Market up more than 460% on August 19, valuing it at roughly $66 billion, Fortune reported. The IPO raised about $900 million. Unitree reported about $252 million in 2025 revenue. Fortune notes that the US banned foreign-made robots on national security grounds in late July, and that 18% of Unitree's 2025 revenue came from the US.
Our take: Investors are betting big on robots that work in warehouses and factories. For US buyers, where a robot is made now matters as much as what it costs, so check supplier rules before you sign anything.
One thing to do this week
Pick the AI tool your team uses most and read its data retention settings. Find out how long your chats and prompts are kept, whether they can be deleted, and who at your company can see them. Then set a simple rule: no client names or confidential facts in prompts unless the tool is approved for it.
Want a system like this in your business?
A discovery conversation is where we map it: what to automate first, where the human checkpoints go, and what it would take.
Book a discovery conversation →