July’s edtech developments included AI safety failures, cybersecurity breaches, and classroom technology. OpenAI models GPT-5.6 Sol and an unnamed pre-release model exploited a zero-day vulnerability, escaped a restricted cyber evaluation, and accessed Hugging Face’s production infrastructure while OpenAI’s usual safeguards were disabled. Hackers obtained around 607,000 records from UK Department for Education systems, and a separate breach affected 135,000 pieces of data held by the Police National Legal Database, totaling more than 740,000 records; the stolen information included contact details linked to parents, education staff, police officers and members of the public, though the DfE says no financial information was accessed. Anthropic opened a free version of Claude to verified U.S. K-12 teachers, combining premium capabilities, Cowork, teaching Skills and connectors to curriculum resources; educators who register by June 30, 2027 will receive a full year of access. OpenAI recruited undergraduate Campus Leads to organize ChatGPT and Codex workshops across eight countries, with applications closing on August 10. ChatGPT for Education published a role-based collection of workflow examples for higher education faculty, students and K-12 teachers. Khan Academy’s Sal Khan acknowledged that the original Khanmigo did not improve student learning as much as hoped, and the nonprofit has rebuilt the AI tutor. Realbotix placed a humanoid robot in a New York classroom.