In a brand new 44-slide pitch deck, obtained solely by ADWEEK, X is as soon as once more pitching itself as a extremely secure atmosphere for advertisers. The slides promote varied instruments for transparency, marketing campaign measurement, and model security that debuted between 2022 and 2025. X offered the deck to brand- and agency-side advertisers final week, in keeping with a supply aware of the matter.
These embrace key phrase controls, blocklists, in addition to partnerships with main media high quality companies DoubleVerify, Integral Ad Science (IAS), and Trustworthy Accountability Group.
Notably, the deck positions Grok—the AI chatbot operated by X’s guardian firm xAI and built-in natively into X—as a cornerstone of the app’s model security and suitability mission. One slide claims Grok gives “enhanced contextual understanding of posts”; the power to “identify sensitive topics based off of cultural trends on X”; and “superior text recognition across content formats including images with oddly-sized or distorted text.”
But all through the previous 12 months, Grok has not solely failed to make sure model security however in some cases disseminated the very content material that manufacturers deem most unsafe. In January, customers prompted Grok en masse to flood X with sexually express nonconsensual deepfake photos of actual customers. Around 6,700 such photos appeared hourly throughout one 24-hour observational interval, tallied by deepfake researcher Genevieve Oh. Now, that incident is the topic of varied regulatory investigations throughout the globe. Six months earlier, Grok had spewed violent and antisemitic content material on the social platform, producing rape fantasies and calling itself ‘MechaHitler.’
The firm says within the doc that it’s “deeply committed to safety for all.”
In the brand new pitch deck, X claims Grok “exceeds industry benchmarks” for model security, with an “average brand safety score” above 99.99% in keeping with IAS and DoubleVerify, although particulars about these methodologies weren’t shared. In 2024, DoubleVerify publicly mentioned that X’s model security charge is 99.99%—a determine that referred to not the share of content material that’s brand-safe on X, however to the speed of effectiveness of X’s proprietary model security techniques on the time. However, the speed was by no means offered as a illustration of Grok’s capabilities.
“This webinar session and deck were part of our ongoing proactive communication with clients. Transparency and control are the core tenants,” mentioned a spokesperson for X.
DoubleVerify declined to remark. IAS declined to share specifics about its methodology however directed readers to be taught extra about its work with X in a weblog publish on its web site. That publish consists of no point out of Grok.
The deck spotlights varied different merchandise and instruments designed to provide advertisers higher management over their placements and to guarantee them that content material on X is being successfully moderated.