Goldman Sachs is pulling Claude Code from its bankers and traders

Not long ago, non-technical staff at Goldman Sachs were using AI coding agents to build their own custom trackers and market models without waiting on software teams. Now, the firm is restricting tools like Claude Code strictly to dedicated engineers to better manage security and data risks on sensitive machines.

This shift highlights an interesting tension in the world of AI building. On one side, giving domain experts the power to create their own software solves problems fast. On the other side, agents that can automatically write and run code introduce real safety questions when deployed in high-stakes environments.

For anyone learning to build with AI, this news shows just how capable these tools have become for everyday problem-solving. As you build your own apps, keeping basic safety in mind—like protecting private data and reviewing what an agent runs—will help you build with confidence.

Community discussion: https://news.ycombinator.com/item?id=50415628