{
  "id": 332780,
  "title": "Anthropic says auto mode will be the default in Claude Code for Pro, Max, Team plans, starting on Aug. 14, claiming it's good enough at catching harmful actions (Simon Willison/Simon Willison's Weblog)",
  "url": "https://urgent.news/2026/08/09/anthropic-says-auto-mode-will-be-the-default-in-claude-code-for-pro",
  "topic": "ai",
  "section": "AI",
  "published": "2026-08-09T00:38:10.000Z",
  "source": {
    "name": "Techmeme",
    "slug": "techmeme",
    "url": "https://simonwillison.net/2026/Aug/8/auto-mode/"
  },
  "original_language": "en",
  "account": "Anthropic has announced that auto mode will become the default setting for Claude Code in Pro, Max, and Team plans, starting on August 14th. According to Simon Willison, Anthropic is highly confident in the auto mode's safety, stating that almost every person in their organization uses auto mode. They claim that Claude Code's auto mode effectively mitigates most risks, such as prompt injection and data exfiltration, with only a small percentage of cases remaining unprotected. However, Willison acknowledges that there are still two safety issues to address: accidental damaging actions and prompt injection. He highlights Anthropic's evaluation by Trajectory Labs, which tested various Claude Code models, including Claude Fable 5, Opus 5, and Sonnet 5, and found that none of the 720 attack attempts succeeded when running auto mode. Willison remains skeptical and calls for more independent confirmation of Anthropic's claims.",
  "summary": "Anthropic is making auto mode the default setting for new sessions in Claude Code for Pro, Max, and Team plans, starting August 14th. According to Simon Willison, Anthropic claims that auto mode is good enough at catching harmful actions. This change was discussed in a Fireside Chat with Cat Wu and Thariq Shihipar at the AI Engineer World’s Fair last month.\n\nResearchers from York University and the University of Calgary found that developers have concerns about AI coding tools, including security and privacy issues. According to Gias Uddin, associate professor at York University, these tools are still relatively new and evolving rapidly, which creates pressure to add new capabilities. The researchers analyzed Reddit discussions and found that many reported issues come from how these tools are designed and what access they are given, not just the underlying models.\n\nThe study's findings suggest that the builders of AI coding tools failed to prioritize security and privacy, leaving developers to defend themselves. The researchers recommend building security and privacy mechanisms into the design before giving a tool broad access to a developer's files, data, or systems.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 2,
    "also_reported_by": [
      {
        "outlet": "Mashable",
        "title": "Lifetime access to Claude, GPT, Gemini, and 17+ other AI models is $60 with this promo code",
        "url": "https://urgent.news/2026/08/08/lifetime-access-to-claude-gpt-gemini-and-17-other-ai-models-is-60",
        "published": "2026-08-08T09:00:00.000Z"
      }
    ]
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}