Anthropic joins White House pledge for AI education

Anthropic has formalized its participation in the White House's "Pledge to America's Youth: Investing in AI Education" by announcing three major initiatives. The AI developer is investing $1 million over three years to fund PicoCTF, a cybersecurity education program run by Carnegie Mellon University for middle and high school students. Additionally, the company is supporting the newly launched Presidential AI Challenge and releasing a free, Creative Commons-licensed "AI Fluency" curriculum designed to help K-12 and higher education instructors integrate AI tools into their classrooms regardless of school resources.

These efforts build on Anthropic's existing footprint in the education sector. The company's Claude model currently powers MagicSchool, a platform serving 3 million educators across all U.S. school districts with over 100 million AI engagements. Claude also supports Amira Learning's reading tools and Solvely.ai, which assists more than 5 million students across 120-plus countries. Meanwhile, the Claude for Education program serves over 100,000 students and faculty, including rollouts at USF Law and Northeastern University's 13 campuses, while Claude Builder Clubs are launching at more than 60 U.S. universities.

To guide these initiatives, Anthropic analyzed over 1 million university student conversations and 74,000 educator conversations. The study revealed that students use AI for complex tasks like creating and analyzing 70 percent of the time, while 57 percent of educator usage focuses on curriculum development and building custom tools with Claude Artifacts. These insights led to the creation of Claude's "Learning Mode" and informed the new fluency curriculum. For AI practitioners and developers, Anthropic's deep integration into schools highlights a growing market for educational technology built on large language models.

Separately, Anthropic announced a $5 million grant program to fund independent research into how AI impacts user wellbeing. The company is also updating its Claude Fable 5 model, modifying its biology safeguards to reduce false positives and minimize the frequency of "fallbacks" to less capable models when users input biology-related queries.