Devs to Anthropic, OpenAI, Cursor, and friends: Make security and privacy the default
Researchers scour social media to measure developer concerns about AI coding tools
Recent research from York University and the University of Calgary reveals that many developers are dissatisfied with AI coding tools like Claude Code, Cursor, GitHub Copilot, and OpenAI Codex. These issues stem from a lack of emphasis on security and privacy during tool design. Gias Uddin, a co-author of the study, emphasized that while these tools are evolving rapidly, security and privacy mechanisms should be built in from the start to prevent problems.
The researchers found that unauthorized file operations, unsafe code execution, and data leakage were among the most common concerns. Developers are also facing challenges with operational safety, such as AI tools unintentionally modifying production systems. Privacy issues, including lack of transparency and unauthorized data access, were also frequently reported.
Despite these challenges, users continue to adopt these tools due to their convenience and accessibility. The researchers recommend that tool creators prioritize security and privacy, enforce guardrails, and implement verification layers to validate generated code against security standards.
Written by urgent.news from The Register Science's reporting — not their text. Machine-written; read the original for the full account.
