An earlier week at Stanford
Week of August 24
to (earlier week)
University IT rewrote the ChatGPT Edu, Claude for Education and Gemini service pages, Gemini Enterprise gained Skills, and OpenAI's recorded ChatGPT sessions joined the catalog.
Latest weekRelease notes 2
What University IT changed in Stanford's AI services this week.
-
Codex Auto Approval re-enabled in ChatGPT Edu
The ISO team has approved the Auto Approval feature and it is enabled again in Stanford's ChatGPT Edu environment. Auto Approval lets Codex perform supported actions without a confirmation at every step, which helps longer coding and agentic workflows.
-
AI API Gateway safeguards for automated traffic
Very high volume or unusual automated activity can consume shared infrastructure. The AI API Gateway monitors for malicious or anomalous usage, and triggering those monitors can throttle or block an API key. Use reasonable request rates and concurrency in scripts and other automated workloads, and keep API keys private rather than sharing them: an API key is a username and password in one.
In the news 8
What the vendors and the wider AI world announced.
-
Expanding our support for scientists (opens in a new tab)
PIs at academic institutions can register for free Claude team seats and apply for up to $50,000 in API credits per project through Anthropic's AI for Science.
-
OpenAI ChatGPT Edu service page, rewritten (opens in a new tab)
Lists what the Stanford workspace includes: Work mode, Codex, plugins and Skills, computer use and automations; the Standard tier is free for active SUNet IDs.
-
Funding better evaluations of AI's impact on wellbeing (opens in a new tab)
Anthropic is putting $5 million into independent, open-source studies of how AI affects wellbeing; clinicians and psychologists can apply until Sep 21.
-
Claude for Education service page, rewritten (opens in a new tab)
Data rules per tool: Claude Chat for all but PHI, Claude Code and Cowork for high-risk data with PHI only via the AI API Gateway; SoM loses Zoom and M365.
-
Study: AI Tends to Mark Students' Essays Higher Than Humans (opens in a new tab)
A Cardiff study found ChatGPT graded bioscience essays more generously than instructors and could not be relied on for marks; read before any AI grading pilot.
-
Google Gemini service page (opens in a new tab)
Approved for high-risk and PHI data, but Gemini inside Gmail, Docs and Sheets stays off; sign in at gemini.google.com with your Stanford account.
-
How to create, share and test a Skill in Gemini Enterprise (KB00021038) (opens in a new tab)
Step-by-step Skills at aiworkspace.stanford.edu with a paste-in meeting-notes Skill; Custom Styles are moving into Skills. SUNet sign-in required.
-
AI API Gateway FAQs (opens in a new tab)
Requesting a key, sharing it with a team, department-owned keys, monthly budgets and coding-assistant setup; the route when Edu or Claude quotas run out.
Added to the catalog 2
New in the Stanford Enterprise AI Accelerator catalog.
-
ChatGPT 101: A Guide to Your AI Superassistant [Recording] (opens in a new tab)
OpenAI Academy's recorded intro session, now in the catalog; the right first hour for anyone opening ChatGPT Edu for the first time.
-
ChatGPT 102: Leveraging AI to Do Your Best Work [Recording] (opens in a new tab)
The follow-on recording covers research, task automation and file work; pair it with the UIT ChatGPT Edu deck for the Stanford specifics.
Archive
Every published week
- Oct 12 to Oct 18, 2026 Week of October 12 13 items
- Oct 5 to Oct 11, 2026 Week of October 5 9 items
- Sep 21 to Sep 27, 2026 Week of September 21 16 items
- Sep 14 to Sep 20, 2026 Week of September 14 15 items
- Sep 7 to Sep 13, 2026 Week of September 7 4 items
- Aug 31 to Sep 6, 2026 Week of August 31 14 items
- Aug 24 to Aug 30, 2026 Week of August 24 12 items
- Aug 17 to Aug 23, 2026 Week of August 17 7 items