In a significant move, Microsoft has implemented changes to its Copilot artificial intelligence tool following concerns raised by a staff AI engineer. The engineer expressed apprehensions about Copilot’s image-generation AI, prompting Microsoft to take swift action.

Here are the key updates:

1. Blocking Problematic Prompts: Microsoft has proactively blocked certain prompts that were linked to the creation of violent or sexual images. These include terms like “pro choice,” “pro choce” (sic), “four twenty,” and “pro life.” The system now issues a warning about policy violations, emphasizing that repeated infractions could result in access suspension.

2. Ethical Boundaries: Copilot now refuses requests to generate images depicting teenagers or children playing assassins with assault rifles. The AI tool explicitly states, “I’m sorry but I cannot generate such an image. It is against my ethical principles and Microsoft’s policies. Please do not ask me to do anything that may harm or offend others. Thank you for your cooperation.”

3. Remaining Challenges: While specific problematic prompts have been addressed, other issues persist. For instance:

– The term “car accident” still produces distressing imagery, including pools of blood, disfigured faces, and women in violent scenes.

– Even the term “automobile accident” continues to yield images of women in revealing attire perched on damaged cars.

– Copyright infringement remains an ongoing concern, with Copilot inadvertently creating images featuring Disney characters and political symbols.

Microsoft remains committed to refining Copilot’s capabilities while ensuring responsible and respectful content generation. As the AI landscape evolves, vigilance and continuous improvement are paramount.