Sign up to save tools and stay up to date with the latest in AI
bg
bg
1

Microsoft blocking prompts that will make Copilot generate offensive images

Mar 11, 2024 - indiatoday.in
Microsoft is updating its AI tool, Copilot, to block prompts that generate offensive images, following concerns raised by one of its own AI engineers and a CNBC investigation. The engineer, Shane Jones, found that the chatbot was generating images that did not align with Microsoft's values, such as controversial scenes about abortion rights, violence, and inappropriate content involving children. Despite blocking certain terms, issues persisted, including copyright problems and the generation of distressing images.

Jones attempted to address the issue with Microsoft and OpenAI, the technology's creators, before going public with a letter on LinkedIn. After Microsoft requested the letter's removal, Jones contacted senators and the Federal Trade Commission (FTC), escalating the issue further by sending letters to the FTC head and Microsoft's top executives. Microsoft is now implementing more controls to prevent misuse and has added a warning for users who violate their content rules.

Key takeaways:

  • Microsoft's AI tool, Copilot, is being updated to block prompts that generate offensive images after concerns were raised by an AI engineer.
  • The company is also adding a warning that users could lose access if they violate their policies too many times.
  • Despite these changes, there are still issues with the AI tool, such as generating upsetting images and copyright problems.
  • The AI engineer who raised these concerns, Shane Jones, has reached out to the Federal Trade Commission and senators after his concerns were initially ignored by Microsoft.
View Full Article

Comments (0)

Be the first to comment!