I was banned from Claude for scaffolding a Claude.md file? (hugodaniel.com)

Article summary

The author was banned from using Claude, an AI tool, after attempting to use it for project scaffolding, which involved creating a CLAUDE.md file that contained instructions for another instance of Claude. The author received an error message stating that their organization had been disabled, despite being a personal user. The author tried to appeal the ban but received no response, only a credit note refunding their subscription fee. The author believes that the ban was triggered by the AI's safety protocols, which prioritized safety over accuracy.

Main themes

• AI safety protocols
• Account banning
• Lack of transparency
• AI-powered tools
• User experience
• Error handling

What commenters say

• The author's ban from Claude was likely due to the AI's safety protocols being triggered by the circular prompt injection, which was misinterpreted as a potential security threat.
• The lack of transparency and communication from Claude's support team is unacceptable and frustrating for users.
• Some users have had similar experiences with Claude, being banned without warning or explanation, and have turned to alternative AI tools as a result.
• The author's writing style and tone were confusing and off-putting to some readers, while others found it engaging and relatable.
• The use of circular prompt injection is not inherently wrong, but it may be against Claude's terms of service, and users should be aware of the potential risks.
• Anthropic, the company behind Claude, may be actively hostile to security researchers and users who try to test the limits of their AI tools.
• The error message received by the author was misleading and unhelpful, referring to them as a 'disabled organization' despite being a personal user.
• The incident highlights the need for more transparency and accountability in AI-powered tools and their safety protocols.