• Sources: primary, discussion
  • Summary: State of AI, citing Bloomberg News, reports that internal Goldman Sachs guidance updated in October restricts Anthropic's Claude Code and Microsoft's GitHub to engineers and strats after a review of productivity and risk. Traders and bankers had used the tools to monitor client flows, model positions and build material, and the firm's in-house multi-model assistant stays available to all staff. The Hacker News thread reached 386 points and 237 comments.
  • Why it matters: Large firms are starting to scope local coding agents by role, so agent sandboxing and ownership of agent-built code become rollout prerequisites outside engineering.
  • Comments: Commenters frame the move as the lethal trifecta of private data, untrusted input and network egress on the same workstation, and note that corporate desktops often block the virtualization that would sandbox agents. Others report companies clawing back Claude access over billing, since commenters say organizations above 150 seats pay API rates rather than flat subscriptions.