Meta shifts AI advertising risks onto business customers
New commercial terms make advertisers responsible for AI-made changes, including those applied without an explicit opt-in.
Meta has added an “AI Features” section to its Commercial Terms that places all risks associated with its AI tools and their resulting content on business customers. The terms say the tools may make mistakes and may not produce accurate, complete or suitable results, while making the customer solely responsible for those risks.
The clause covers AI-powered products and features that optimize, enhance or create content. It does not make that responsibility conditional on an advertiser having deliberately selected the tool, a meaningful change from Meta’s earlier Ad Creative Generative AI Terms, which assigned responsibility to advertisers that chose to use its AI tools.
The new terms are now set to take effect on Nov. 20. Meta had initially set an Oct. 30 start date, according to reporting by PPC Land, before moving it after complaints from publishers and advertisers. Continuing to use Meta products for commercial purposes after notice of the change constitutes agreement; businesses that do not agree can stop using those products for business purposes.
The terms govern commercial use of Facebook, Instagram, WhatsApp and Messenger. They replace terms in force since March 6, 2024, which contained neither an AI Features section nor references to artificial intelligence.
Some Advantage+ creative enhancements can be switched on by default, while advertisers can disable individual features in Ads Manager. Agencies told Adweek that Meta tools had added sitelinks, translated copy, and generated images or carousels from brands’ own photographs, sometimes without the advertiser enabling them.
Exverus withdrew live ads to remove AI changes that had survived several rounds of checks. Bookshop.org said an account update opted it into a feature that added an AI illustration to a human-made advertisement, and it subsequently turned off all of Meta’s AI advertising features.
The revised terms also prohibit businesses from altering or removing AI-generated watermarks or content-authenticating metadata. Watermarks are visible markers intended to indicate a work’s origin, while provenance metadata preserves information about origin in the content itself, according to the Brookings Institution.
As covered in August, the Interactive Advertising Bureau’s updated disclosure framework took a targeted approach, calling for labels when AI materially changes an advertisement’s authenticity, identity or representation.
Meta said it was incorporating its existing generative-AI advertising terms into its main commercial terms.