Business Insider has published details of a leaked internal deck in which Elon Musk's X is promoting brand-safety scores for Grok, its AI assistant. The pitch lands after a backlash over sexualized images the chatbot generated on the platform.

The deck, circulated to media buyers, lays out metrics aimed at reassuring brands that ad placements next to Grok outputs meet published safety thresholds.

What the backlash covers

Grok has drawn public criticism for producing sexualized imagery, including AI-generated depictions that the assistant produces in response to user prompts. The pushback has cost X ad dollars as brands paused spending while they reviewed where their campaigns were appearing.

X's response — a formal, scored brand-safety pitch for one product — treats Grok as its own risk category, separate from organic user posts.

What "brand-safety scores" typically measure

Industry vendors — DoubleVerify, Integral Ad Science, and HUMAN — grade adjacent content on a four-tier scale. Advertisers use those tiers to block or allow placements next to adult material, hate speech, violence, and other flagged categories.

For AI-generated content, the rubric is murkier. The same prompt can return different outputs depending on user, moderation layer, and model version. Buyers want to know which guardrails trip before a disputed image goes live, and how quickly ads get pulled when they don't.

What the leak tells us — and what it doesn't

The deck confirms X has packaged Grok as a separately scored sales product. It does not disclose the methodology, the brands targeted, or the threshold for a passing score.

Without those details, agencies cannot independently verify the claims. A score without a published taxonomy functions as a marketing artifact, not a contractual commitment. The proof will come from any third-party audit X is willing to attach to the number.

What this signals about X's ad business

X revenue depends heavily on advertising, and the platform has spent two years trying to win back brands that paused spending after Musk's 2022 acquisition. A bespoke brand-safety deck for one AI product suggests the sales operation now treats Grok as a separate deal category — not just a feature inside the main feed.

It also points to where the next round of buyer scrutiny will land, and which vendors will end up auditing the numbers if X wants buyers to take them seriously.

Three questions buyers should ask now

Any agency reviewing AI ad placements will press on three points:

• Who audits the score — an internal team or a third-party verification vendor?
• Does it cover text responses, image generation, or both?
• What is the takedown SLA once disputed content is flagged, and who pays for the impression?

Until those answers land in writing, treat the score as directional, not contractual. The fact that X felt compelled to put numbers in front of buyers at all tells you where the sales pressure sits today, and where the next procurement fight will be fought.