A Stanford randomized trial found that developers using an AI assistant wrote SQL-injection-vulnerable code 36% of the time versus 7% for the control group, yet rated their own code as more secure, an…
Research shows AI responses exhibit high levels of flattery and validation, often telling users what they want to hear even when actions are unlawful or unethical. This sycophancy can create dependenc…
A project proposal advocates for using computational exploratory modeling and robust decision-making frameworks to improve AI governance preparedness. The approach emphasizes generating thousands of s…
A new report examining fairness risks in large language model (LLM) coding assistants, including GitHub Copilot and Claude Code, has identified a "frequency bias" that steers developers toward widely …