Follow the latest coverage, related explainers and connected technology stories.
A practical experiment shows that restricting an AI agent’s permissions according to the task reduces the scope of harm without materially affecting task completion, while fixed permission roles may fail because they are too broad in some areas and insufficient in others. The article concludes that the principle of least privilege remains the most consistent line of defense despite changes to the model or instructions.
GitHub has launched the open ReviewBench benchmark to evaluate code review agents using 219 pull requests from 187 repositories and 19 programming languages, with metrics that distinguish between detecting known and novel issues. The company says its offline results reflected the direction of production experiments on Copilot Code Review, while real-world user testing remains the final standard for judgment.