Microsoft Copilot is introducing a new safeguard to keep AI-generated content professional and respectful. This update adds a bad language filter designed to prevent offensive or inappropriate terms from appearing in Copilot outputs. It’s part of Microsoft’s ongoing commitment to responsible AI and fostering inclusive digital workspaces.

What’s New?

Here’s a quick look at the key improvements:

• Language Moderation: Copilot now automatically detects and filters inappropriate or offensive language in generated content.
• Cross-App Support: The filter works across Microsoft 365 apps, including Word, Outlook, and Teams.
• Consistent Tone: Helps maintain professional communication standards in documents, emails, and chats.
•  

Impact on Users

No disruption to existing workflows—Copilot continues to deliver intelligent suggestions, now with added safeguards for tone and language.

What You Need to Do

No action is required. The filter is enabled by default and applies automatically to Copilot-generated content.
Admins can review and manage harmful content protection settings if needed:
Manage harmful content protection settings for Microsoft 365 Copilot

Compliance

This feature supports organisational compliance and inclusivity goals. No additional compliance concerns have been identified, but organisations should review internal policies as needed.

How Harmful Content Protection Works in Copilot Chat

Policy Control

When an organisation applies a policy to adjust AI protections, users will see a Harmful Content Protection toggle in the More menu of Copilot Chat.

Default Setting

Harmful Content Protection is enabled by default at the start of every conversation, this means Copilot will block harmful or offensive content.

Disabling Protection

• Users can switch the toggle off to disable harmful content protection for the current conversation only.
• When disabled, queries may return sensitive or potentially offensive content.

Important Limitation

• Once disabled, the setting cannot be turned back on during the same conversation.
• To re-enable protection, start a new conversation.
•