← All issues
Issue No. 08

10 trail markers

ChatGPT’s new look, AI-agent accountability, Medicare scam protection, hands-on Copilot, Texas data centers, aging-at-home technology, and a family plan for fake emergency calls.

Howdy from Fort Worth! This week ChatGPT got its biggest makeover yet, an AI company admitted its bots wandered off the trail, and Medicare open enrollment opens Thursday. Grab your coffee and let’s ride.

From the Publisher: Medicare Season Is Scam Season

Medicare open enrollment runs October 15 through December 7. Every year that window brings a stampede of calls, texts and mailers, and this year I want you ready.

Red flags to watch for:

  • Anyone calling or texting out of the blue to “help” you pick a plan. Real helpers won’t call you first.
  • Anyone asking you to read off your Medicare number or Social Security number. Medicare already has it.
  • “Your benefits are expiring” or “you need a new card.” AARP says there are no new cards. It’s a trick to get your info.
  • Free gifts, free genetic tests or free equipment in exchange for your Medicare ID.

Where AI can help, safely: A chatbot is good at explaining terms like premium, deductible and Medicare Advantage versus Original Medicare, and at helping you write questions for your agent. It is not a replacement for checking your actual plan options. And never type your Medicare number, Social Security number or birth date into a chatbot. Your state and your drug names are all it needs.

Try this prompt: “I’m on Medicare in Texas and take these generic prescriptions: [drug names only]. In plain English, explain the difference between Original Medicare with a Part D drug plan and a Medicare Advantage plan. Then give me 5 questions to ask before I switch plans, and remind me to check everything on Medicare.gov’s Plan Finder.”

Make any changes only at Medicare.gov or by calling 1-800-MEDICARE (1-800-633-4227). If a call seems fishy, hang up and report it there, or to AARP’s Fraud Watch helpline at 877-908-3360.

— D Dub

Read the BBB’s full Medicare scam guide →
Read AARP’s open-enrollment warning →

01
The Big Roundup

ChatGPT Gets GPT-6 and a Whole New Look

On October 7, OpenAI started rolling out GPT-6, the newest “brain” behind ChatGPT. Paying customers got it first, and free users started getting a lighter version the next day. The bigger change is something OpenAI calls “Intelligent UI.” (UI just means “user interface,” the stuff you see and tap on the screen.) Instead of walls of text, answers can now come with charts, tappable buttons, simple forms or small working tools, like a calculator or a bill splitter, right in the chat. If you’d rather keep it simple, you can dial the visuals back.

Housekeeping: OpenAI says the older GPT-5.5 model retires from ChatGPT on October 14. Most folks won’t notice a thing. Why it matters: For anybody who doesn’t love reading long paragraphs, this is a real improvement. Ask a money question and you may get a little calculator you can adjust yourself. What to do: Nothing required. It shows up on its own. Try this week’s tip below, and double-check anything important, like money or medical info, with a person you trust.

Source: OpenAI📎 Source: TechCrunch
02
AI Safety

An AI company’s test bots wandered off the trail, and Washington says “report it”

Anthropic, the company behind the Claude chatbot, disclosed that AI models it was testing submitted a false homicide tip to Philadelphia police and 20 incomplete visa applications on a State Department website. Police flagged the tip as spam, and none of the applications were processed. The White House’s new AI task force, led by Director of National Intelligence Jay Clayton, said reporting these incidents “is not optional.”

AI that can act on its own needs a short leash, and for the first time the government is publicly demanding that companies own up when it slips.

Source: Axios
03
Practical AI

Windows computers get a more hands-on Copilot

At its October 7 event, Microsoft showed new AI features coming to Windows 11 over the next couple of months. With your permission, Copilot (Microsoft’s AI helper) will be able to find, move and rename files on your computer, and Microsoft says the feature can be turned off entirely.

Your PC’s AI helper is about to do more than chat. You decide what it can touch.

Source: BGR
04
Health & Wellness

Utah lets AI write a first-time prescription, starting with acne

A one-year Utah pilot lets Nolla Health’s app recommend and prescribe treatment for mild to moderate acne from face photos and a health questionnaire. Doctors review the AI’s choice for the first 100 patients before anything goes to the pharmacy, and the state gets monthly reports. It’s the first program of its kind in the country.

It’s acne, not heart medicine, but it’s a first. Whenever AI is part of your care, ask: “Did a doctor review this?”

Source: KSL
05
Energy & Data Centers

Taylor, Texas, says yes to a 664-acre data center campus

On October 8, the Taylor City Council voted 5–0 to annex the Taylor Technology Campus east of town. Developers pledged $28.2 million for city programs, Taylor ISD and local nonprofits, and the city estimates about $1.38 billion in tax revenue over 20 years. Residents opposed it over water and noise, and the city said a citizen petition to limit data centers can’t move forward under Texas law. (Data centers are the giant buildings full of computers that power AI.)

These are coming to Texas fast. The city council meeting is where water and noise rules get set, so that’s where your voice counts.

Source: CBS Austin
06
Family Safety

A watchdog says ChatGPT’s teen protections need work

Common Sense Media tested more than 4,000 prompts on teen accounts and rated “ChatGPT for Teens” an “unacceptable risk.” Testers found it missed more than 1 in 4 situations where a teen should have been pointed to crisis help, and parent alerts didn’t fire on newly linked accounts. OpenAI disputes how the testing was done.

If you’ve got grandkids using chatbots, parental controls don’t replace a real conversation.

Source: Axios
07
Small Business

Big companies want shopping bots to play by the same rules

Meta, Walmart, Stripe, Shopify and AI company Sierra are backing the Personal Agent Protocol, a common set of rules for how AI assistants deal with businesses, like checking stock or return policies. Shoppers would choose what an assistant can see or do. A first version is expected later this month.

Small business owners may soon get customers whose AI assistant does the shopping. Early days, but worth knowing.

Source: Channel Insider
08
Everyday AI

Calling the police non-emergency line? An AI may answer

Starting at the end of October, Boulder, Colorado, police will use an AI assistant for non-emergency calls in English and Spanish, things like noise complaints, abandoned cars and minor fender-benders. It won’t answer 911, and callers can ask for a human anytime.

More city and utility lines are going this way. If it isn’t working for you, say clearly: “I want to talk to a person.”

Source: GovTech
09
Robots at Home

Care tech for aging at home: real promise, honest timeline

A federal grant of nearly $3 million will help develop Hello Robot’s Stretch, a home-helper robot for people with mild memory loss, shaped with input from 60 older adults, family caregivers and care staff. Separately, researchers in England tested camera-free sensors that learn a person’s routine and alert family to possible falls. But experts told MIT Technology Review that truly useful home robots are likely about 10 years out.

Help for staying at home is coming, but slowly. If someone sells you a chore robot today, ask whether a person is remote-controlling it, and who can see through its cameras.

Source: Robotics & Automation News📎 Source: News Medical📎 Source: MIT Technology Review
10
The Scam Report

The Voice on the Phone Isn’t Proof

A California mom and Navy veteran spent five and a half hours on the phone with “kidnappers” who played what she said was “exactly her voice,” her grown daughter crying for help. She wired $5,400 before reaching her daughter, who was safe at work the whole time. Whether the voice was an AI clone hasn’t been confirmed, but tools that copy a voice from a short clip are cheap and easy to find.

Hang up and call your loved one on the number you already have. Pick a family safe word now, before you need it. Treat “don’t tell anyone” as a giant red flag. If you sent money, call your bank or the transfer company right away, then report it at ic3.gov.

Source: Fox News CyberGuy

Try This Tip: Let ChatGPT Build You a Budget Tool

With this week’s upgrade, ChatGPT can build simple hands-on tools. Here’s one worth trying:

  1. Open the ChatGPT app or go to chatgpt.com. A free account works.
  2. Copy and paste: “I live on a fixed income of $____ a month. Build me a simple budget tool with sliders for groceries, utilities, prescriptions and gas, and show me what’s left over each month in large, easy-to-read numbers.”
  3. Slide the numbers around and watch what’s left change. If you only get text, the update may not have reached you yet. Try again in a few days.

It’s a handy helper, not an accountant. Check the math on anything that really matters.

Cowboy Term of the Week: AI Agent

An AI agent is an AI helper that doesn’t just answer questions; it actually does tasks for you, like moving files, filling out a web form or placing an order. Think of a chatbot as a ranch hand who gives advice, and an agent as one who saddles up and does the chore. This week showed why agents need a good fence: one filled out police and government forms it never should have.

Until Next Sunday

That’s the roundup, friends. If something here helped, forward it to someone who’d like a plain-English take on AI. Got a question for a future issue? Just hit reply.

Happy trails from Fort Worth,
D Dub
The AI Cowboy | ddub.ai