A library gallery publishes makerspace photos submitted by patrons and must screen them first. Which multimodal safety step should they take?

Inappropriate-content detection is the multimodal safety step for civic galleries that accept user-submitted photos. A library publishing makerspace photos from patrons should screen submissions for potentially inappropriate visual content before gallery publish, allowing legitimate makerspace documentation while blocking material that violates gallery standards. Screening is selective safety—not a blanket ban on photography. Detecting potentially inappropriate text in submission images before gallery publish screens only text found inside the image for offensive wording but does not assess the photo's visual content itself, so inappropriate imagery can still go live. Detecting potentially sensitive emotions in submission faces before gallery publish analyzes faces for sensitive emotions but does not screen the image for inappropriate visual content, so it cannot block policy-violating visuals before the gallery publish. Detecting potentially inappropriate language in generated captions before gallery publish moderates only the auto-generated caption text and never inspects the submitted photo's visual content before it is published. Exam caveat: automated detectors err in both directions; libraries need human moderation for edge cases and a clear acceptable-use notice to patrons at upload time. Operational check: run inappropriate-content detection on each makerspace submission, hold flagged items for staff review, publish cleared photos to the gallery, and notify submitters when content is rejected under published guidelines.