TikTok proclaims new watermarking plans for AI-generated content material

TikTok is making one other try to curb the rising unfold of AI-generated misinformation on social media platforms by asserting a brand new strategy Watermarking (or labeling) AI.

The platform introduced At present it joins the Adobe-led firm Content material Authenticity Initiative and that Coalition for Content material Provenance and Authenticity (C2PA)a nonprofit undertaking designed to coach expertise and social media corporations on greatest practices for what’s referred to as “content material provenance,” or the “fundamental, trusted info a few digital piece of content material’s provenance.”

Whereas TikTok already labels AI-generated content material (AIGC) created utilizing its personal AI results and requires customers to label their very own uploaded AI content material, the brand new coverage will apply computerized monitoring to content material created outdoors the location . TikTok’s said intent with the collaboration is for an auto-tagging system to learn content material credentials to research a picture or video’s metadata and rapidly determine it as AI-generated.

The platform additionally introduced that it will start attaching content material credentials to TikTok content material itself in order that others may be taught when, the place and the way the content material was created or edited. It’s now the primary social media firm and video platform to signal on Adobe content material credentials requirements, Belongings Stories.

“Whereas consultants typically suggest AIGC labeling to help accountable content material creation, additionally they warn that labels may cause confusion if viewers lack context about what they imply,” the corporate wrote in its announcement. “That’s why we labored with consultants to develop media literacy campaigns that may assist our group determine and suppose critically about AIGC and misinformation.”

TikTok may also launch 12 new media literacy assets created along with these from the Poynter Institute Media-relateda youth-focused fact-checking undertaking, and WITNESS, a human rights group that teaches civilians use expertise to file and shield themselves. WITNESS additionally gives pointers and Recommendation for stylish deepfakes and different AI threats.

Different expertise platforms, additionally flooded with generative AI merchandise, have taken their very own paths towards higher labeling and watermarking. Meta introduced new AI content material labels In April. Not lengthy after, Snapchat introduced that it will add an computerized, however user-visible, watermark to all content material created with its in-house AI instruments. The clear Snapchat emblem watermark is added to pictures as they’re downloaded to a tool or exported from the platform.

In the meantime, TikTok is rolling out these meant AI protections as consideration is drawn to the platform’s function in international organizing and its potential influence on the upcoming presidential election. The brand new insurance policies additionally come amid the corporate’s last-ditch efforts to safe the app’s place on U.S. telephones, together with a current lawsuit towards the U.S. authorities over its Biden-approved ban.