Ask The PhD Community

HigherEd AI Daily

July 2: Export Controls Lift as AI Research Tools Expand

Thursday, July 2, 2026

Federal regulators reversed an AI export freeze this week while Anthropic rolled out a faster model and a dedicated research workbench for scientists.

TLDR AI | GOVERNANCE

Washington Reverses Export Ban on Anthropic's Advanced AI Models

The Department of Commerce ordered Anthropic on June 12 to suspend access to its Claude Fable 5 and Mythos 5 models for all foreign nationals, citing national security concerns after a trusted partner identified a jailbreak vulnerability; Anthropic maintained that the workaround was minor and comparable to weaknesses found in other publicly available models. After an 18 day standoff and a security review, Commerce lifted the export controls on June 30, restoring license free export, reexport, and in country transfer of both models.

Anthropic says it has implemented new safeguards targeting the specific vulnerability that triggered the freeze. The episode shows how quickly federal agencies are willing to apply export control authority to consumer facing AI products, not just chips or hardware.

Why it matters for campuses

International students, visiting scholars, and cross border research partnerships depend on uninterrupted access to frontier AI tools. A sudden foreign national access freeze, even a temporary one, can disrupt coursework, grant funded research, and lab collaborations without warning. Institutions with international campuses or exchange programs should treat this as a preview of how national security reviews can reach into everyday campus AI use, and should have contingency plans ready for tool access interruptions.

Read More

TLDR AI | TOOLS

Anthropic Ships Claude Sonnet 5, Narrowing the Price Gap to Flagship Performance

Anthropic released Claude Sonnet 5, a mid tier model built for stronger agentic performance in planning, tool use, coding, and knowledge work; the company says its capabilities approach the flagship Opus 4.8 while representing a substantial improvement over Sonnet 4.6. Sonnet 5 is now the default model for Free and Pro plans and is also available to Max, Team, and Enterprise users, as well as through Claude Code and the Claude Platform, with introductory API pricing of $2 per million input tokens and $10 per million output tokens through August 31, after which prices rise to $3 and $15.

Anthropic also reports that safety assessments found Sonnet 5 shows a lower rate of undesirable behaviors than its predecessor and is generally safer to use in agentic contexts, a relevant data point for institutions evaluating which models to approve for student and staff use.

Why it matters for campuses

A cheaper model with near flagship performance changes the cost calculus for campus wide AI deployments, from writing centers to research computing. Instructional designers and IT leaders weighing which tier of Claude to standardize on now have a materially stronger low cost option, and the improved agentic reliability may ease concerns about deploying AI assistants in advising or administrative workflows.

Read More

TLDR AI | RESEARCH

Claude Science Consolidates Research Tools Into a Single AI Workbench

Anthropic introduced Claude Science, a beta app for Pro, Max, Team, and Enterprise users on macOS and Linux that brings the tools researchers already rely on into one environment; it natively renders 3D protein structures, genome browser tracks, and chemical structures, and connects to more than 60 scientific databases with prebuilt toolkits for genomics, proteomics, structural biology, and cheminformatics.

Early adopters report substantial time savings; a researcher at the Allen Institute used Claude Science to build an automated pipeline for literature reviews that previously took up to two years, completing reviews exceeding 100 pages. Anthropic is also funding up to 50 Claude Science research projects with as much as $30,000 in credits each, with applications open through July 15.

Why it matters for campuses

This tool is aimed squarely at faculty and graduate researchers in the sciences, and the funded project program gives research offices a concrete, low cost opportunity to get involved. Research computing and library staff should evaluate whether Claude Science fits existing data governance and compute access policies before recommending it broadly, particularly for work involving controlled or sensitive datasets.

Read More

TLDR AI | RESEARCH

What AI's Uneven Progress in Mathematics Signals for Teaching and Research

In a wide ranging conversation with podcaster Dwarkesh Patel, 3Blue1Brown creator Grant Sanderson examines how AI models are advancing unevenly across mathematics; systems can now brute force certain areas, such as geometry, with striking speed, but still struggle with combinatorics problems that require the playful, exploratory creativity that experienced mathematicians describe as central to discovery.

Sanderson argues this unevenness previews how AI will likely transform other knowledge domains: rapid gains in well structured problem spaces, slower gains where judgment, taste, and creative leaps matter most, and a continued need for human curation to make sense of AI generated work.

Why it matters for campuses

Math, computer science, and STEM departments are already contending with AI tools that solve homework style problems instantly. This conversation gives faculty a more precise vocabulary for where AI genuinely displaces routine problem solving versus where it still depends on human framing and judgment, which is useful for redesigning assessments and research training around the reasoning skills students still need to be taught.

Read More

Tool of the Day

Nano Banana 2 Lite

Google's fastest and most cost efficient text to image model, generating a 1K resolution image in about four seconds for roughly $0.034 each. It is available now in Google AI Studio, the Gemini API, and rolling out across the Gemini app and NotebookLM, giving educators an inexpensive, fast way to produce original visuals without design software.

Try it: Generate three variations of a diagram illustrating a core concept from an upcoming lecture, then compare which version best supports student comprehension before adding it to your slides.

Visit Nano Banana 2 Lite

Have a great learning day!

Dr. Ali Green

Sources for This Edition

TLDR AI (tldr.tech)
CNBC (cnbc.com)
Anthropic (anthropic.com)
Dwarkesh Podcast (dwarkesh.com)
Google (blog.google)

askthephd.com
 | 
askthephd.substack.com
 | 
Unsubscribe

HigherEd AI Daily; Curated by Dr. Ali Green