What's New in Your AI Tools
Zoom AI Companion
- Zoom's My Notes bot can now cover meetings you can't attend — Zoom AI can join Google Meet and Microsoft Teams calls on your behalf and deliver the AI notes, summary, and action items alongside your regular Zoom meeting notes. It requires calendar sync, and admins can enable or restrict the setting at the account, group, or user level. Useful for the recurring cross-platform meeting that never comes with notes. Also new on Zoom's What's New page: ZM+ for ZoomMate runs Anthropic models inside Zoom's trust boundary, so select AI capabilities work without your data leaving Zoom.
Microsoft Copilot with Data Protection
- Copilot Chat adds text selection and shareable sessions — Microsoft's August roundup rolled out two practical chat upgrades. You can now select a sentence, paragraph, or table inside a Copilot response and ask Copilot to work on just that content, with the selection previewed in the prompt before you send it. You can also share a full chat session or a single response through a link; recipients get a read-only view and can continue the conversation within organizational controls.
Copilot for Microsoft 365
- Cowork gets effort levels and clearer cost controls — Copilot Cowork now lets you choose an effort level, from Light for simple tasks up to Max for deeper analysis, with the model and effort controls sitting together in the compose box. The /cost skill now shows credits used in the current session, the share of your monthly limit remaining, and when the limit resets. The Scheduled tab has been renamed Automations and now covers event-triggered tasks as well as scheduled ones. Also in August: Word gains hyperlinks, image understanding, and highlights; Excel adds chat history and a change-history skill; and Teams can generate meeting recaps in other languages.
Google Gemini & NotebookLM
- Custom instructions for Gemini now work across more Workspace apps — The persistent instructions feature that launched in Docs earlier this year now extends to Ask Gemini in Drive, Ask Gemini in Chat, and the Gemini side panel in Gmail, Sheets, and Slides. Set your tone, length, and formatting preferences once and Gemini applies them across all of those surfaces, so you can stop re-explaining your preferences in every prompt. The gradual rollout began September 2.
- Google Pics brings AI image creation and editing to Workspace — Google Pics is generally available as of September 4, adding AI image generation and object-based editing directly into Workspace creative workflows, so images for newsletters and decks no longer require a separate tool. Also new this week: Google Vids can turn Docs, PDFs, and Word files into narrated video summaries, with AI-generated scripts and custom visuals.
Coming Up: Trainings & Workshops
- The Secret Life of LLMs: The Hidden World of Experts Shaping AI — Wednesday, September 16, 1–2 p.m. PT, virtual. The UC San Diego AI Alumni Council hosts leaders from Handshake AI and Databricks for a look at what happens after model training: who actually teaches and evaluates frontier models, and why human expertise has become one of AI's most valuable inputs. Registration required.
- 2026 AI and Healthy Aging Symposium — Thursday, September 24, 9:30 a.m.–4 p.m. at Park & Market, hosted by the Stein Institute for Research on Aging. Keynotes from Noom's chief medical officer and UC San Diego's Amy Sitapati, plus expert panels on independent living, personalized healthcare, and using new technologies safely. Free to attend, but space is limited and registration is open now.
Self-paced options in the meantime:
- Take the AI Foundations course to learn core AI concepts and UC policies on AI tools.
- Watch the AI Webinar #6 recording — a practical walkthrough of TritonGPT's first year, including MyDocuments, model switching, and chat sharing.
- Explore the Everyday I AI video series for short, practical prompting tips that work across TritonGPT and other AI tools.
TritonAI News
- TritonGPT 4.6.6 reaches production — TritonGPT was upgraded to version 4.6.6 in production this week. The default model now runs Muse Glimmer 30B, an open model hosted on UC San Diego infrastructure, and the lineup adds GPT-5.6 Terra, Gemini 3.7 Flash, and Claude Sonnet 5. Voice Mode lets you speak with an assistant and hear its responses, a coding sandbox runs code inside chat, and reasoning effort is adjustable per model. Read the release details on the TritonAI site.
- SCOUT is in production beta at UC Merced — UC Merced's SCOUT assistant is in production beta, serving more than 100 faculty and staff users on the shared TritonAI platform, with the same hosting and governance controls as the UC San Diego deployment. The expansion sits alongside BearGPT at UC Berkeley, ANR AI at UC Agriculture and Natural Resources, and FredGPT at Fairleigh Dickinson University.
- UC San Diego will lead the new NSF NAIRR Operations Center — NSF awarded the San Diego Supercomputer Center $35 million over five years, with UT Austin's Texas Advanced Computing Center as partner, to establish and run the National Artificial Intelligence Research Resource Operations Center. The center transitions NAIRR from a successful pilot into sustained national infrastructure; the pilot has already connected roughly 900 research teams and educators across all 50 states with AI computing, data, and tools.
- First dual-use incubator cohort includes AI and autonomy founders — The Office of Innovation and Commercialization and Bow Capital selected six companies founded by UC San Diego students, faculty, and alumni for the inaugural summer incubator at the Design and Innovation Building. Each team receives $100,000 and three months of full-time support to build dual-use technologies spanning medical devices, advanced materials, artificial intelligence, and autonomous systems.
Have feedback on this newsletter? Get in touch.