Advice on the Use of AI-Generated Media Content at UiB

The guidelines apply to all media content (text, audio, images, and video) published under UiB’s logo and brand. This includes content on UiB’s own platforms, as well as content published on external platforms such as social media, streaming platforms, printed publications, and presentations. 

You are responsible for staying updated on UiB’s guidelines when publishing under the UiB brand. As a public institution, UiB should not challenge ethical boundaries, but act with caution. 

Main Principles

• We must not present something as real, authentic, or self-produced if it is not.

• We must not produce or publish content that, individually or over time, may weaken trust in UiB as a sender.

• We must make it easy for recipients to exercise critical source evaluation.

• We must manage UiB’s trust and credibility with accuracy, openness, and clarity. 

Caution, Credibility and Source Criticism 

UiB must help ensure that it is easy for audiences to assess sources critically. Content (images, video, text, audio) may be shared beyond our control indefinitely. 

We must label content when necessary to avoid doubt about its origin, authenticity, or purpose. 

Be aware of limitations in generative AI tools, including potential bias. We should not rely on generative AI for precise or accurate information and must treat it critically like any other source. 

We are responsible for ensuring that all published content is correct and up to date, regardless of whether AI has been used.

Personal Data

• Avoid using personal data in open or unsecured AI services.
• Be certain what constitutes personal data in the material you process.
• An image is considered personal data if individuals can be identified.
• Audio recordings may also constitute personal data, even if no names are mentioned.
• Individuals are responsible for their own personal data used in such services.
• Tools used must comply with UiB requirements for data protection and privacy.

Guidelines for Use of Generative AI in Content Production

We may:

• Use AI for text-related tasks such as proofreading, translation, transcription, generating titles, summaries, ideas, metadata, research, and brainstorming.

• Publish AI-generated text, but only after human editing, quality assurance, and fact-checking. You are responsible for content quality. 

• Use AI to improve workflows in editing photos, video, and audio (e.g., noise reduction, resizing, trimming, adjusting light/contrast).

• Use AI-generated visual content similarly to illustrations, drawings, graphics, or animations, provided professional quality is ensured and copyright considerations are respected.

• Use AI for audio editing and production under the same principles as for other media types. 

We may not:

• Publish AI-generated text without human quality assurance.

• Present AI-generated text as fully self-produced work — its use must be disclosed.

• Publish AI-generated visual content as if it depicts real events, places, or people.

• Publish manipulated images or video in ways that could harm UiB’s credibility.

• Manipulation includes altering content so that its meaning or context changes or misrepresents reality (e.g., adding nonexistent people or objects).

• Publish AI-generated audio that could be mistaken for real individuals (e.g., voice clones) without clearly labeling it. 

Underlying Principles 

Images and video have traditionally been regarded as highly credible evidence, but AI makes manipulation easy and often indistinguishable from reality. 

Graphics, drawings, and animations do not pose the same ethical risks, as they are not typically mistaken for reality—but transparency is still essential. 

Voices are also strong identifiers and have traditionally been trusted. AI now makes it harder to distinguish real from synthetic audio, increasing the importance of source criticism. 

UiB must not present anything as real, authentic, or self-produced if it is not. Transparency and integrity are essential to maintain credibility and enable critical evaluation. 

UiB is responsible for the overall impression created under its brand, regardless of medium or platform. 

How to Label AI-Generated Content

• Audio in videos or podcasts can be labeled:
  • On end screens
  • As tags in the video
  • Verbally in the audio track
  • Or in accompanying text on relevant platforms
• Images and video should be labeled directly on the content so the information follows the media. 
• For AI-generated text, the use of AI must be clearly stated so there is no doubt about the extent of AI involvement.