[May 2026 Update] Following Anthropic securing SpaceX's Colossus-1 data center (220,000+ GPUs), Claude Code rate limits were doubled across all paid plans (Pro/Max/Team/Enterprise) and peak-hour restrictions were removed—reducing the congestion-based slowdowns previously experienced by heavy users.

For the past several months, developers running Claude Code daily on Pro had been hitting an invisible wall—sluggish responses after 5 PM, and throttled requests toward month's end. Before Anthropic secured the Colossus-1 deal, compute was rationed across users during peak hours, so heavy users took the hit first. That structural bottleneck is now gone.

If you run Claude Code on most workdays—solo developer, freelancer, or daily power user—this is effectively a capacity upgrade at no extra cost. You won't need to ration prompts or bank headroom for end-of-month crunches. For anyone comparing Cursor vs. Claude Code, the value gap just widened. If you open Claude once or twice a week, the change is invisible—you were never hitting the cap.

New liability front: In 2025, law firm Latham & Watkins submitted a brief in Concord Music Group v. Anthropic containing AI-generated hallucinations, raising questions about attorney disciplinary liability for unverified AI output—a distinct risk category from copyright disputes.

Until recently, the dominant AI-legal story was about copyright: whether training on published data constitutes infringement, as argued in NYT v. OpenAI and Getty v. Stability AI. Lawyers were already using AI to draft documents, but the failure mode was mostly embarrassment—a hallucinated citation flagged by opposing counsel. The Latham & Watkins incident shifts that frame: unchecked AI output now carries disciplinary exposure, not just professional embarrassment.

For attorneys, consultants, and anyone who submits professional documents under their name, this raises the bar for AI-assisted work—skipping verification now risks bar complaints or sanctions. The case draws a clear line: AI liability varies heavily by role, and using AI for a court brief is very different from using it for an internal note. Documenting your team's AI review process—what gets checked before anything goes out the door—is worth the afternoon it takes. For those using AI only for internal notes and summaries, the direct exposure remains low for now.