AI developer Anthropic has launched a self-hosted public beta for its coding tool, Claude Code, allowing Team and Enterprise users to run coding sessions on private servers. While built for corporate software development, this release points to a challenge facing school districts: adopting artificial intelligence while protecting sensitive student data. Schools are implementing AI in classrooms, and administrators must weigh convenience against strict privacy laws.
What Happened
Anthropic has moved its self-hosted environments for Claude Code into public beta, according to EdTech Innovation Hub. This feature lets organizations run agent sessions on infrastructure they control, placing Claude Code alongside their internal networks and security tools. These setups rely on customer-managed runners, which are background processes that pull sessions from a queue and run separate isolated processes for each user. Under this model, sensitive code files, secrets, and build results remain entirely on the customer’s private network. However, the system still transmits conversation transcripts and tool results back to Anthropic for processing. Anthropic notes that organizations utilizing zero data retention (ZDR) agreements cannot use this self-hosted feature.
The Bigger Picture
While corporate IT teams manage self-hosted networks, school districts are facing their own data-handling challenges with classroom tools. In July 2026, Anthropic launched Claude for Teachers, a program that offers free premium AI services to verified K-12 educators in the United States. This tool encourages teachers to upload student rosters and lesson diagnostics, and it integrates with classroom software like Canva Education and MagicSchool.
This setup can create security risks. According to The 74, individual teachers signing up for the tool on their own bypass district security vetting. This creates a "privacy trap" that can violate state and federal laws. Under the Family Educational Rights and Privacy Act (FERPA), software vendors must operate under the direct control of school districts when handling student records. Under the Children’s Online Privacy Protection Act (COPPA), school systems must explicitly disclose any third-party AI subprocessors that process student information.
Beyond privacy, educators must consider how these tools impact learning. A study in Educational Technology Research and Development revealed that students with lower computational thinking skills tend to use AI assistants to copy answers rather than learn. To address this, platforms like zyBooks build custom "AI Hints" designed to offer Socratic guidance rather than immediate solutions, which encourages active learning. This matches research in the Journal of Computers in Education, which confirms that AI adaptive systems only improve student retention when structured around careful, teacher-led guidance. As we have previously noted, defining clear educational outcomes is necessary as these emerging autonomous tools redefine cybersecurity education in classrooms.
What This Means for Families
For parents, these changes show that classroom AI adoption is complicated. When teachers use unauthorized tools, student records and work go directly to technology companies. School districts must establish secure cloud environments with Zero Data Retention (ZDR) agreements to prevent third-party models from storing student data. Anthropic's new feature shows how organizations are often forced to choose between data privacy and access to collaborative AI tools.
What You Can Do
- Ask your school's administration if they have a central vetting process for AI software used by classroom teachers.
- Review your school district's privacy policies to ensure any AI platform used by teachers has a signed FERPA compliance agreement.
- Support the use of scaffolded, school-approved AI platforms like zyBooks that promote problem-solving instead of simple answer-copying.