Why Human Oversight of AI Isn't Always Enough

Why human oversight of AI isn't always enough

One risk of people using genAI tools like ChatGPT for work is that they over rely on what the tools produce. That’s particularly problematic when those tools make errors or make up information entirely, and why many people advocate for “human-in-the-loop” designs, which ensure a person reviews AI work before it’s finalized.But a new working paper highlights the limitations of this approach for large language models (LLMs). Even when people revie…

The Video AI Hate Problem: Why Safety Takes Backseat to Speed — Human in the Loop

In October 2025, researchers at the Anti-Defamation League's Centre on Technology and Society conducted a deceptively simple experiment...