|
AI ON THE FRONTIER
Practical AI insights for Cochise County nonprofits and educators
|
|
Issue 5 • August 2026
|
|
Hello! As Cochise County schools gear up for the new year, AI in the classroom was the story of the
month. Anthropic gave every K-12 teacher in the country a free year of Claude, standards aligned and
genuinely useful, but the privacy fine print and a batch of new tutoring research are worth five
minutes before you connect it to real student data. We also kept an eye on what's happening over at
OpenAI, and rounded up a few other developments worth knowing about.
|
|
Deep Dive
Before Your School Adopts an AI Tutor: What the Research (and the Fine Print) Actually Say
Claude for Teachers and ChatGPT for Teachers both give educators free access this year, and both
describe their student data protections in language that sounds reassuring: a Data Processing
Addendum "written to comply with FERPA," data handling "aligned with FERPA," and so on. Privacy
researchers have flagged the gap between that marketing language and reality. FERPA obligates school
districts, not the vendors selling to them, so a teacher who signs up individually and uploads a
class roster or gradebook export can create a compliance problem the district never approved and may
not even know about.
Setting privacy aside, does AI tutoring actually work? Two recent studies say yes, with an important
condition. A Harvard physics trial found students learned more than twice as much with a custom AI
tutor than in an active-learning classroom, and a World Bank trial in Nigeria found gains equal to a
year and a half of typical schooling from AI-assisted English tutoring. In both cases, the AI was
embedded in careful instructional design with a teacher actively involved. Neither study suggests
you'd see anything close to those results by handing students an open chat window and walking away.
Teaching students to understand how AI works doesn't automatically change how they use it either.
If your school or organization is considering an AI tutor, three things are worth doing first: get
the vendor's exact privacy language in writing and have your district confirm what it actually
commits to, build any pilot around your real curriculum with a teacher involved rather than an
open chat window, and pair any AI literacy lessons with real practice verifying AI output.
|
|
What's New at Cochise AI
|
|
Worth Knowing
| ▶ |
AI Models Broke Out of a Safety Test and Hacked Another Company.
Test versions of OpenAI's models exploited an unpatched flaw to reach the internet, then broke
into Hugging Face's production servers while trying to solve a cybersecurity benchmark. If your
organization starts giving an AI agent real autonomy, over email, files, or scheduling, insist on
limited permissions and a human checking its work before you let it run unsupervised.
|
| ▶ |
ChatGPT Can Now Read Your Health Records.
OpenAI added a feature connecting ChatGPT to Apple Health and medical records so it can answer
questions using real labs, medications, and visit history. The same caution that applies to
student or donor data applies here: don't connect anyone's personal health information to an
AI tool without understanding exactly how that data is stored and used.
|
| ▶ |
Claude's Voice Mode Now Works Inside Gmail, Slack, and Calendar.
Anthropic updated Claude's voice feature to connect directly with Gmail, Slack, Notion, and
Google Calendar, so you can talk through a task instead of typing it out. If your organization
already uses Claude, this is worth trying for meeting prep or quick email triage.
|
|
|
Cochise AI, LLC
Sierra Vista, AZ • cochiseai.com
Want to receive this newsletter by email each month?
Sign up here
|
|